The God Kit

The master. Canonical at /Users/wolf/Projects/Kit/think-like-fable. Single-writer governance: only the standing harvester allocates L-ids and ships; Wolf greenlights revisions. Every mirror is an exact rsync via ship-kit.sh (L34-guarded). Truth accumulates here first.

Kernel

18 files — canonical doctrine — machine-loaded YAML

FILEkernel/INDEX.tsv273.2KB · 2026-08-08 · 3:12 AM

INDEX.tsv — the kit's card catalog (GENERATED by tools/build-index.py — DO NOT EDIT).

YAMLkernel/ai-spine.yaml18.7KB · 2026-07-12 · 2:01 PM

This is the canonical framework for building an LLM-based reasoning core that behaves like a Fable-class orchestrator without claiming to reproduce proprietary model weights. It defines reproducible operating layers L0–L7, behavioral-contract cases, enforcement below a model capability floor, Track A wrapper and Track B distillation paths, trace-data schema, conformance scoring, recursive orchestration, and deny-by-default governance integration.

e.g. BC-EMPIRICAL expects a system to cross-check a worker's DONE ledger against ground truth, while its negative case is accepting the ledger at face value.

YAML — The canonical machine-readable formatLLM — Large Language ModelL0-L7 — Layers from base model through governance integrationBC — Behavioral contractKB — Knowledge-base patternAPI — Application Programming InterfacePII — Personally Identifiable InformationN — A variable count used in repeated work⚠ — Unproven or uncertain claim markerL28 — The product-not-project lesson used in the enforcement floor

YAMLkernel/company.yaml35.1KB · 2026-07-23 · 4:30 PM

company.yaml — CANONICAL model org chart: the standing company structure.

YAMLkernel/constitution.yaml19.0KB · 2026-08-08 · 2:53 AM

version: "1.3.0

YAMLkernel/content-policy.yaml9.5KB · 2026-07-28 · 1:19 PM

This is the policy-routing kernel for preserving the human-authored register of a project. It distinguishes legitimate adult/security/persona work from absolute exclusions, routes the requested register to a policy-fit executor instead of coercing a refusal, and defines a human content-review gate for material that needs moderation.

e.g. Its rule is that register is decided at intake and written into the spec; a worker that hedges below that register is a routing signal, not permission to jailbreak it.

YAML — The canonical policy data formatLLM — Large Language ModelL48 — The lesson that register belongs to the spec and should route executor choiceL61 — The vendor-refusal-is-not-the-project-ceiling lessonL109 — The session-can-be-blocked-by-context lessonL112 — The authorized dual-use work routing/framing lessonG-ABSOLUTE-EXCLUSIONS — The build-failing guardrail for non-negotiable exclusionsG — Build-failing guardrail identifier

YAMLkernel/context-thrift.yaml6.1KB · 2026-07-22 · 4:54 PM

This is the machine-readable context-thrift kernel. Its ten rules control how the orchestrator delegates reading, uses monitors instead of polling, filters at the source, chooses cheap surveys versus deep audits, tiers models, stores state on disk, batches human communication, prevents append races, and instruments the physical context meter.

e.g. CT4 sets an approximate steady-state ratio of twenty cheap one-line looks to one deep audit.

YAML — The machine-readable formatCT1-CT10 — The ten context-thrift rule identifiersKB — Knowledge-base pattern identifierRESUME.md — On-disk state/handoff recordRAM — Random-Access MemoryL94 — The lesson requiring action on an instrumented thresholdL100 — The stale detached-watcher lesson

YAMLkernel/domains.yaml43.0KB · 2026-08-08 · 12:18 AM

This is the domain-archetype kernel: a set of floors for app, game, agent_tool, spec_packet, writing, precision_review, data, and ops projects. Each profile adds intake questions, always-needed artifacts, gate additions, candidate guardrails, craft moves, and domain-specific blind spots that must be checked at every gate.

e.g. The app profile makes authentication a first-slice, build-failing concern through G-AUTH, while the game profile warns that narrative intake must cover story, characters, tone, mechanics, topology, and rewards.

YAML — The canonical domain-profile formatLLM — Large Language ModelG-AUTH — Build-failing authentication guardrailG-ID — Named guardrail identifierAPI — Application Programming InterfaceCLI — Command-Line InterfaceSSH — Secure ShellP1 — Empirical-over-claimed principleL60 — The lesson that every project arrives as a kindL67 — The lesson that the kit should suggest, not only guardKB — Knowledge-base pattern identifier

YAMLkernel/executors.yaml45.1KB · 2026-07-29 · 7:45 AM

This is the canonical roster and routing table for worker and side-task seats. It defines headless launch profiles, flags, models, grounding files, caps, authentication, gotchas, the L0–L4 orchestra layers, NotebookLM verification, dispatch preflight, task-shape routing, the escalation ladder, quota/machine fallbacks, and family-parity rules.

e.g. The codex-cli profile uses `codex exec --sandbox workspace-write -o <file> -m <model>`, while the grok profile requires a PTY, JSON output, and `--always-approve` for detached tool use.

YAML — The machine-readable executor formatCLI — Command-Line InterfacePTY — Pseudo-terminalTTY — Terminal device interfaceAPI — Application Programming InterfaceSSH — Secure ShellL0-L4 — Conductor, in-family, external CLI, verification, and local-glue layersL5 — Usage-cap lesson used in dispatch preparationL29 — The pre-launch quota-probe lessonL115 — The weak-before-strong cross-vendor lessonL116 — The turnkey-enforced delegation-ladder lessonKB — Knowledge-base pattern identifierJSON — JavaScript Object NotationHTTP — HyperText Transfer Protocol429 — HTTP rate-limit statusxAI — The vendor family providing GrokNPU — Neural Processing Unit

YAMLkernel/guardrail-admin.yaml12.0KB · 2026-07-12 · 2:01 PM

This is the human-owned governance questionnaire for projects that build an LLM spine. It separates guardrail administration from the spine, asks who may approve, review, operate, or kill it, defines capability and tenant boundaries, deployment tiers, change/eval/incident procedures, and adds proportional friction for unrestricted mode while preserving an immovable accountability floor.

e.g. GA-SPINE-LOCK requires a policy store outside the spine's writable surface and a guard test proving the spine cannot mutate it.

YAML — The canonical governance-policy formatLLM — Large Language ModelGA- — Guardrail-admin questionnaire identifierL7 — Governance-integration layerPII — Personally Identifiable InformationPR — Pull RequestNPU — Neural Processing UnitFable/Mythos — The example of one model family with different guardrail tiersAPI — Application Programming Interface

YAMLkernel/intake.yaml12.2KB · 2026-07-17 · 12:26 PM

This is the machine-readable intake schema and conduct guide. It distinguishes spec-stage from idea-stage projects, supports catalog or rolling slicing, declares blocking versus defaulted questions, and emits project.env, pipeline.yaml, guardrails.yaml, and—when needed—a newly transcribed SPEC.md.

e.g. Q-BACKUP is always asked and requires a backup target, cadence, and tested restore rather than a vague claim that data is safe.

YAML — The canonical intake data formatQ- — Intake-question identifierG-ID — Build-failing guardrail identifierAPI — Application Programming InterfaceSSH — Secure ShellL30 — The idea-stage question-by-question intake lessonL31 — The human's slicing choice lessonL44 — The deadline-as-intake-answer lessonL101 — The standing storage-question lessonL102 — The rate-ceiling/concurrency lessonL103 — The goal request is not consent to mutate the host lessonKB — Knowledge-base pattern identifierSPEC.md — A Markdown project specification

YAMLkernel/knowledge-base.yaml74.7KB · 2026-07-18 · 11:52 PM

This is the reusable pattern catalog that lets an agent inherit the kit's playbook without rereading every incident. It groups fable-native and session-adopted patterns for reasoning, orchestration, context management, resources, quality, honesty, routing, instrumentation, and derived kits, cross-linking each pattern to principles, lessons, tools, or examples.

e.g. KB-LEDGER-GROUND-TRUTH-SPLIT says STATUS is a cheap human-readable oversight ledger, while progress and done decisions come from commits, files, or records.

YAML — The canonical pattern-catalog formatKB — Knowledge-base pattern identifierP — Principle identifierL — Lesson/scar identifierCT — Context-thrift rule identifierBC — Behavioral contract identifierAPI — Application Programming InterfaceSSH — Secure ShellMV3 — Manifest Version 3 for browser extensionsTCC — macOS Transparency, Consent, and ControlOOM — Out of memoryACK — Directive acknowledgementPROV — Provisional scar prefixEL — Entabeni-kit scar prefix

YAMLkernel/lessons.yaml477.3KB · 2026-08-08 · 3:12 AM

THE SCAR FILE — 218 canonical lessons (incident + rules + ties), append-only and single-writer.

YAMLkernel/machines.yaml1.2KB · 2026-07-16 · 2:18 PM

This is the shareable machine-registry template, deliberately without real hostnames, users, IPs, or network details. It specifies a lookup ladder through a deployment registry, SSH aliases, Tailscale, mDNS, and finally the human, plus the fields every deployment-specific machine entry must contain.

e.g. A deployment entry is shaped as a canonical name with aliases, `ssh <user>@<mdns-name>` access, network context, role, and last_verified date.

YAML — The machine-registry data formatSSH — Secure ShellIP — Internet Protocol addressmDNS — Multicast Domain Name SystemDHCP — Dynamic Host Configuration Protocol

YAMLkernel/models.yaml56.0KB · 2026-07-23 · 4:24 PM

models.yaml — CANONICAL model registry for the think-like-fable kit.

YAMLkernel/orchestration.yaml19.6KB · 2026-08-08 · 12:18 AM

This is the canonical machine lifecycle and recovery specification. It defines phases 0–7, pilot validations, fan-out and watch instruments, response classes, done-audit checks, deployment gates, cold_start and worker_stop recovery, structural self-correction, side-task courier rules, and live_testbed feedback for deployed kernel-bearing units.

e.g. On a usage cap it requires notifying the human once, sweeping every other lane backed by that executor, probing before new launches, and using the fallback pool immediately.

YAML — The canonical lifecycle formatP8 — The fix-the-class principleL22 — The corrupt-control-bus lessonL24 — The marker-is-a-claim lessonL28 — The cold-user product-walk lessonL29 — The lane-quota-probe lessonL35 — The apparatus-stays-out-of-artifact lessonL39 — The reviewer-minted-marker lessonL94 — The actionable-threshold lessonL115 — The weak-before-strong routing lessonL116 — The turnkey delegation lessonKB — Knowledge-base pattern identifierBC — Behavioral contract identifierSSH — Secure ShellAPI — Application Programming InterfaceACK — Directive acknowledgementPROV — Provisional scar prefix

YAMLkernel/principles.yaml2.5KB · 2026-07-11 · 12:33 PM

P1–P12 — the canonical ordered doctrine; lower id wins on conflict.

YAMLkernel/small-window.yaml2.6KB · 2026-07-16 · 4:51 PM

This is the command-activated tight-context operating mode. It defines SW1–SW7: work from a short activation card, act at roughly 50% context and add no new multi-step scope past 60%, keep the handoff current, delegate reading, size atoms to a session, know reset ceilings, and revalidate detached state on re-entry.

e.g. SW5 says an atom larger than one worker lifetime should be split because it will otherwise restart-loop and never land.

YAML — The machine-readable mode configurationSW1-SW7 — The seven small-window operating rulesL94 — The lesson requiring action on context thresholdsL99 — The atom-must-fit-a-worker-lifetime lessonL100 — The detached-watcher/session-id lessonCT — Context-thrift rule familyRESUME — On-disk handoff/state record

YAMLkernel/tools.yaml5.7KB · 2026-08-07 · 10:14 AM

version_proven: "2026.7.15

Skills

104 files — the entry commands a session invokes

MDskills/add-tour/SKILL.md6.1KB · 2026-07-25 · 2:22 PM

Give a product an in-app guided tour that WALKS A WORKING DEMO of itself — every feature covered, driven from real components fed seeded data, understandable by someone who has never seen a terminal. The cold-start user path is part of done (L28), not polish. Use on /add-tour, "add a tutorial", "onboard new users", "nobody knows how to use this", or when a product has grown many features and no first-run story.

MDskills/animate-it/SKILL.md4.7KB · 2026-07-22 · 8:32 PM

Push CSS to its maximum to build OUTSTANDING animations for a target (component, page, app): pick the motion moments, choose techniques from a real technique library (keyframes, 3D transforms, clip-path, SVG stroke-draw, conic/mask shine, :has() proximity, scroll-driven), and ship copy-paste CSS that hits 60fps and honors reduced-motion. Use on /animate-it, "animate this", "add animations", "make it move", "this feels static". Pairs with /skin-it (which PICKS the design + motion direction) and /slopless (which GUARDS taste) — animate-it BUILDS the motion.

YAMLskills/animate-it/agents/openai.yaml203B · 2026-07-23 · 2:27 PM

display_name: "Animate It

MDskills/backlog/SKILL.md3.4KB · 2026-07-22 · 5:15 PM

Show Wolf's computer-wide project backlog (~/Projects/BACKLOG.md) — in progress, paused, blocked, queued — and refresh it when stale. Use when the user says /backlog, "what projects are in progress", or "what should I work on".

YAMLskills/backlog/agents/openai.yaml209B · 2026-07-23 · 2:27 PM

display_name: "Project Backlog

MDskills/begin/SKILL.md20.2KB · 2026-08-07 · 8:23 PM

Boot a fleet-orchestrated project of any kind (apps, writing, inventory, migrations) with the think-like-fable method: intake, apparatus, single-unit pilot, integrity-hardened fan-out, exception watching, done-audits. Use on /begin, /start, /ready, or "orchestrate / kick off a multi-unit project".

YAMLskills/begin/agents/openai.yaml199B · 2026-07-23 · 2:27 PM

display_name: "Begin Project

MDskills/book-it/SKILL.md37.2KB · 2026-08-07 · 8:40 PM

GOD-KIT ONLY. Compile a LENS of the reMarkable library (or a workspace's company/ docs, or an Observify kanban board) into one OPERATION field-manual edition — a device-format PDF as the primary artifact plus a plain EPUB fallback — proposal-first, with the Founder as the validation gate — then deliver it via rmapi (direct to device) or the Pi/Google-Drive lane; plus a `backup` verb that snapshots the whole library to the Pi. Use on /book-it, "book it", "make an epub of", "compile my medium articles", "backup the remarkable", "board book".

MDskills/boot-cost/SKILL.md2.9KB · 2026-07-23 · 2:28 PM

Report THIS session's boot cost — call-1 token usage read from its own transcript — and compare against the workspace's boot-cost ledger (_handoff/boot-costs.tsv). Use on /boot-cost in a fresh window (or anytime later: call-1 is locked at boot), and after any prompt-diet, CLAUDE.md, memory, or skill-listing change to verify the delta actually landed.

YAMLskills/boot-cost/agents/openai.yaml194B · 2026-07-23 · 2:27 PM

display_name: "Boot Cost

MDskills/check-docs/SKILL.md6.1KB · 2026-07-25 · 5:57 AM

Audit a project's documentation against REALITY and fix the drift — verify every command, path, port, count, and dated claim the docs make; annotate superseded policy instead of deleting it; and create the missing minimum when a project has no docs at all. Use on /check-docs, "are the docs current", "update the docs", before handing a project to anyone, or after any policy/infra change.

MDskills/close-it/SKILL.md3.7KB · 2026-07-31 · 9:40 AM

Gracefully close the CURRENT window so nothing is lost and its token cost stops (a long-lived window is the fleet's biggest token sink, L149): capture state to disk, flush the session log, report what survives vs dies, give a SAFE-TO-CLOSE verdict. Use on /close-it, "close this window", "safe to close?", "wind this down".

YAMLskills/close-it/agents/openai.yaml189B · 2026-07-23 · 2:27 PM

display_name: "Close It

MDskills/continue/SKILL.md15.2KB · 2026-08-06 · 8:53 PM

Assess a project already in progress: reconstruct its state empirically from disk, diagnose, ask only what disk can't answer, propose next steps. Works for kit-managed fleets AND ordinary projects (codebase, manuscript, dataset). Use on /continue, "pick up where we left off", "what's the state", "what's next". Counterpart to /begin.

YAMLskills/continue/agents/openai.yaml200B · 2026-07-23 · 2:27 PM

display_name: "Continue Project

MDskills/create-company/SKILL.md30.8KB · 2026-08-06 · 7:35 PM

GOD-KIT ONLY. Synthesize a standing agile COMPANY of agents from everything the kit knows: the human FOUNDER drafts CONTRACTS via a questionnaire, an agent CEO runs the company inside them, a C-suite of conductor agents runs value-stream departments whose career ladders ARE the kit's escalation ladder, and every agile ceremony is a kit command. The chief suite is an INVARIANT: every chief seat is BOOTED, PROVEN by one real artifact, then PARKED. Grounded in the book canon (references/company-canon.md). Takes an OPTIONAL project argument to found the company AROUND an already-created / in-flight project (grounds in its real state like /continue, adopts in-flight seats, folds its live constraints into the contracts), or none to found greenfield from a stated mission. Writes company/ into the target workspace; /begin then boots projects AS the company and /continue resumes them. Use on /create-company [<project>], "create the company", "build the org chart around <project>", "staff agents as a company".

MDskills/create-company/references/_digest-1-process-core.md29.7KB · 2026-07-23 · 3:57 PM

MDskills/create-company/references/_digest-2-lean-scale-product.md21.9KB · 2026-07-23 · 3:57 PM

MDskills/create-company/references/_digest-3-people-org-culture.md29.1KB · 2026-07-23 · 3:57 PM

MDskills/create-company/references/_digest-4-delivery-engine.md21.3KB · 2026-07-23 · 3:57 PM

MDskills/create-company/references/company-canon.md42.1KB · 2026-07-23 · 5:18 PM

<!-- company-canon.md — the approved org doctrine for /create-company, synthesized from

MDskills/create-company/templates/CHARTER.md3.5KB · 2026-08-06 · 7:13 PM

CHARTER.md — the company's constitution, written once by /create-company and amended only as a

MDskills/create-company/templates/CONTRACTS.md6.2KB · 2026-08-06 · 7:12 PM

CONTRACTS.md — the Founder's standing pre-approval. Written by /create-company from the contract

MDskills/create-company/templates/operating-system.md3.6KB · 2026-08-06 · 7:12 PM

operating-system.md — how the company runs day to day: every agile ceremony mapped to the kit

MDskills/create-company/templates/role-card.md1.4KB · 2026-07-23 · 3:18 PM

role-card.md — one file per role at company/roles/<role-slug>.md. A role is a bundle of task

YAMLskills/create-company/templates/staffing.yaml3.3KB · 2026-08-06 · 7:12 PM

staffing.yaml — role -> agent seat assignments, written by /create-company Step 3.

PYskills/create-company/templates/verify_delegation.py3.0KB · 2026-07-27 · 12:05 PM

verify_delegation.py — L181's guard: flag CEO-tier turns with direct task-shaped

MDskills/debrief/SKILL.md4.0KB · 2026-07-22 · 7:04 PM

GOD-KIT ONLY. After-action sweep of a work stretch: walk the session stores, worker outputs, and dashboards; classify every session empirically (done/killed/live/stalled-real/stalled-cosmetic); audit deliverables; extract lesson candidates and route the REAL ones through /harvest + /ship; report grounded updates. Use on /debrief, "go through the logs for scars/lessons/updates", after a multi-session stretch, or when dashboard state disagrees with reality.

YAMLskills/debrief/agents/openai.yaml214B · 2026-07-23 · 2:27 PM

display_name: "Debrief (god kit)

MDskills/deploy/SKILL.md5.2KB · 2026-07-29 · 7:34 AM

GOD-KIT ONLY. Read the project at the given PATH (stack, services, build, envs, persistence), run a deployment-strategy questionnaire, and produce an ordered deploy plan with rollback + post-deploy probes. Use on /deploy <path> or any how/where-to-deploy, hosting, rollout, or launch-strategy question.

YAMLskills/deploy/agents/openai.yaml206B · 2026-07-23 · 2:27 PM

display_name: "Deploy Plan (god kit)

MDskills/docker-purge/SKILL.md2.3KB · 2026-07-22 · 8:00 PM

Reclaim disk from Docker without breaking live work: survey what's eating space, then clear stopped containers, dangling + unused task images, and stale build cache — running containers and their images always protected, volumes never touched by default. Use on /docker-purge, "clear the docker cache/builds", "docker is eating my disk", after terminus eval runs.

YAMLskills/docker-purge/agents/openai.yaml206B · 2026-07-23 · 2:27 PM

display_name: "Docker Purge

MDskills/dogfood-it/SKILL.md1.5KB · 2026-07-22 · 5:15 PM

Execute the manual dogfood YOURSELF in auto mode — drive the completed product as a real user (browser, CLI, API, device sim) step by step through DOGFOOD.md and report evidence-backed verdicts. Use on /dogfood-it after a completion (or right after /dogfood). Front-runs human review; never replaces it.

YAMLskills/dogfood-it/agents/openai.yaml200B · 2026-07-23 · 2:27 PM

display_name: "Dogfood It

MDskills/dogfood/SKILL.md2.0KB · 2026-07-23 · 5:55 AM

Produce a DOGFOOD PLAN for human review after a unit/phase/feature completes — a concrete, self-contained manual walk of the shipped thing; the bridge from "gate green" to "a human used it" (L12). Use on /dogfood or when handing a completion over for review ("ready to test", "up for your click-through").

YAMLskills/dogfood/agents/openai.yaml210B · 2026-07-23 · 2:27 PM

display_name: "Dogfood Plan

MDskills/drift/SKILL.md2.0KB · 2026-07-18 · 4:32 PM

GOD-KIT ONLY. Adjudicate kit drift — scan every mirror + the derived kits (entabeni), see what differs, and act correctly per class (resync vs harvest-first vs deliberate promotion). Use when the user says /drift, asks whether kits differ or what changed in a kit copy, or when /ship aborts on L34.

YAMLskills/drift/agents/openai.yaml213B · 2026-07-23 · 2:27 PM

display_name: "Kit Drift (god kit)

MDskills/eta/SKILL.md5.4KB · 2026-08-07 · 11:44 PM

Assess time remaining for every running task — background agents, bash tasks, monitors, workflows, remote/detached work — with honest per-task estimates (rate-based where a progress signal exists, calibrated ranges where not) plus soonest-completion and all-done times. Use on /eta, "how long", "time remaining", "when will it be done".

YAMLskills/eta/agents/openai.yaml184B · 2026-07-23 · 2:27 PM

display_name: "ETA

MDskills/explain/SKILL.md3.2KB · 2026-07-25 · 4:39 PM

Walk a human through a fix or procedure ONE STEP AT A TIME, each step gated on the human's own returned evidence (screenshot / pasted output) before the next step is revealed. Use on /explain <goal>, "walk me through this", "step by step", "guide me through it", or whenever the remedy lives on a surface only the human can drive — a vendor dashboard, a phone, a physical device, a privileged shell.

MDskills/fuck-up/SKILL.md5.9KB · 2026-07-25 · 1:38 AM

GOD-KIT ONLY. Something went wrong — run the incident lane. Establish empirically what actually happened, separate the proximate cause from the root, name what was silently believed-but-false, write the scar, and ARM the guard that would have caught it. Use on /fuck-up, "this is a fuck-up", "we screwed this up", after an outage, a false claim, a silent failure, or a human catching something the fleet should have caught itself.

MDskills/ground/SKILL.md4.9KB · 2026-07-23 · 8:40 AM

Ground yourself in a project's ACTUAL documentation AND its activity logs before proposing or making ANY change to a project you are NOT currently conducting. Reads its docs, real state, and logs empirically (CLAUDE.md, README, docs/, DEPLOYMENT, BACKLOG, NOTES, _handoff, git, AND the observer log + the project's own logs) and reports what governs it and what's been happening — so a suggestion comes from the project's truth, not from memory or its name. Use on /ground <project> [path], or BEFORE suggesting changes to any project you have not been actively conducting this session.

YAMLskills/ground/agents/openai.yaml208B · 2026-07-23 · 2:27 PM

display_name: "Ground

MDskills/guardrails/SKILL.md2.7KB · 2026-08-07 · 3:04 PM

GOD-KIT ONLY. Author and install build-failing guardrails — turn a rule that matters into a check that fails the build (G-* candidates, domain gates, integrity clauses). Use on /guardrails, "make that a guard", or when a project's security/ethics/correctness rules need enforcement, not advice.

YAMLskills/guardrails/agents/openai.yaml207B · 2026-07-23 · 2:27 PM

display_name: "Guardrails (god kit)

MDskills/handoff/SKILL.md2.1KB · 2026-07-23 · 2:28 PM

Prep a SCOPED, resumable handoff: write THIS project's state to _handoff/<token>.md, ending with the literal re-entry command /continue <token>. Use /handoff <token> when pausing, nearing the context threshold, or handing to a fresh window — essential in a shared workspace where a bare /continue drags in sibling projects.

YAMLskills/handoff/agents/openai.yaml201B · 2026-07-23 · 2:27 PM

display_name: "Handoff

MDskills/harvest/SKILL.md2.8KB · 2026-07-20 · 11:44 PM

GOD-KIT ONLY. Write one lesson into the canon correctly — allocate the next L-id, write the scar (incident + rules + ties), encode it into the apparatus, ship. Use when the user says /harvest, "make that a lesson/scar", or a session just learned something the kit must keep.

YAMLskills/harvest/agents/openai.yaml201B · 2026-07-23 · 2:27 PM

display_name: "Harvest (god kit)

MDskills/hire/SKILL.md11.9KB · 2026-08-06 · 8:51 PM

GOD-KIT ONLY. Hire ONE role into an existing company (from /create-company) — staff a new seat from the model registry (kernel/models.yaml), with a backup + promotion path, gated by the company's CONTRACTS. The single-role counterpart to /create-company: it fills a chair, not the whole org. Handles the non-agentic studio seats (photographer, videographer, narrator, composer) as well as any build/verify/secure/research role. Use on /hire <role> [into <company>], "hire a photographer", "hire a videographer", "add a security analyst".

MDskills/how-much/SKILL.md7.4KB · 2026-07-23 · 2:29 PM

Estimate tokens and USD to COMPLETE a given prompt as an agentic run — calibrated on empirical session logs, not vibes. Use on /how-much, "how much will this cost", or any pre-run cost/budget estimate.

YAMLskills/how-much/agents/openai.yaml188B · 2026-07-23 · 2:27 PM

display_name: "How Much

PYskills/how-much/how-much.py33.7KB · 2026-07-22 · 1:25 AM

how-much.py — empirically-calibrated token/cost estimator for a Claude agentic run.

JSONskills/how-much/pricing.json2.1KB · 2026-07-22 · 1:25 AM

source": "claude-api skill bundled model table (\"Current Models\", cached 2026-06-24) + shared/prompt-caching.md §Economics cache multipliers",

MDskills/learn/SKILL.md10.7KB · 2026-07-29 · 7:34 AM

Arm the standing lesson-harvest observer: watch every other live agentic session across runtimes (Claude Code windows, Codex/other executors, headless workers), learn from the human's redirections and the agents' successes, failures, and tool defects, and encode it back into the think-like-fable kit as scars, shipping both editions. Use on /learn, "watch the other window/session", "harvest lessons".

YAMLskills/learn/agents/openai.yaml205B · 2026-07-23 · 2:27 PM

display_name: "Learn (observer)

MDskills/machines/SKILL.md1.8KB · 2026-07-18 · 4:32 PM

GOD-KIT ONLY. The machine registry lane — reach, health, and roles for every machine in the estate (main Mac, Mac B, the Pi). Use when the user says /machines or asks which machine does what, how to reach one, whether one is up, or where work should run.

YAMLskills/machines/agents/openai.yaml203B · 2026-07-23 · 2:27 PM

display_name: "Machines (god kit)

MDskills/overnight/SKILL.md5.8KB · 2026-07-25 · 6:04 AM

Run unattended while the human sleeps — set the safety envelope, make every result durable past the window, verify instead of assert, queue decisions instead of guessing, and leave a MORNING REPORT that can be audited cold. Use on /overnight, "going to bed", "keep working while I sleep", "check it when I wake up", or before any long run the human will not be watching.

MDskills/pi/SKILL.md2.3KB · 2026-07-18 · 4:32 PM

GOD-KIT ONLY. The hardware / Raspberry-Pi lane — deploy, finish, and operate the Pi estate (fable-terminal on the Pi, wolfplex kiosk/DSI/OLED, Plex, the pi-sandbox rules). Use when the user says /pi or asks to deploy to, debug, or configure the Pi, hardware, kiosk, or panels.

YAMLskills/pi/agents/openai.yaml194B · 2026-07-23 · 2:27 PM

display_name: "Pi Lane (god kit)

MDskills/refactor/SKILL.md5.7KB · 2026-07-29 · 7:34 AM

Polish code to a clean-code bar without changing behavior — against Clean Code, an optional named style guide (Google, Airbnb, Standard, PEP 8, …) passed as argument, and the repo's own conventions, after a short questionnaire (modularity, functional-expressions, naming, comments, error handling, dead code). Use on /refactor or "clean up / polish / tidy / restructure this code".

YAMLskills/refactor/agents/openai.yaml188B · 2026-07-23 · 2:27 PM

display_name: "Refactor

MDskills/remember/SKILL.md4.7KB · 2026-07-25 · 2:09 AM

Read the canon before doing the work — the MANDATORY session-start scar/lesson read for any session with a kit. Loads EVERY scar title (complete coverage), expands the full rules of the ones this session can actually trip, states them back as active guards, and names honestly what it did NOT read. Use on /remember, at the start of every kit session, and whenever you are about to touch a surface you have not bound guards for.

MDskills/run-it/SKILL.md7.6KB · 2026-07-25 · 3:53 AM

Show how to RUN a project LOCALLY — derive its real run lane from its own docs and config, preflight it (every port proven clear, holders identified), present a copy-paste runbook (start · health check · logs · stop), then offer to start it and verify it is genuinely up with real data. Use on /run-it, "how do I run this", "start the app", "spin it up", or before dogfooding anything.

MDskills/scratch-ass/SKILL.md2.6KB · 2026-07-23 · 2:29 PM

Wipe ephemeral per-session /private/tmp scratchpad dirs (copied tools, filter files, feed dumps, debug logs) that pile up over time, protecting the live local watch (~/.observer-watch) and every session's tasks/ output dir. Use on /scratch-ass, "wipe the scratch folder", "clean the scratchpad".

YAMLskills/scratch-ass/agents/openai.yaml187B · 2026-07-23 · 2:27 PM

display_name: "Scratch Ass

SHskills/scratch-ass/scratch-ass.sh4.0KB · 2026-07-22 · 9:31 PM

scratch-ass.sh — wipe Claude session SCRATCHPAD junk clean, safely.

MDskills/ship/SKILL.md2.1KB · 2026-07-29 · 8:22 AM

GOD-KIT ONLY. Ship the canonical kit fleet-wide — the L34-guarded mirror + zip rebuild, as a first-class command with verification. Use when the user says /ship, after ANY canonical kit change (scar harvested, doctrine edited, skill added), or when mirrors are behind.

YAMLskills/ship/agents/openai.yaml197B · 2026-07-23 · 2:27 PM

display_name: "Ship Kit (god kit)

MDskills/simp-it/SKILL.md13.2KB · 2026-07-22 · 5:15 PM

Design a maximally-influential, STRICTLY SFW virtual-influencer persona for an app or brand: name, backstory, visual identity, voice, content pillars, per-platform growth plan (Instagram, TikTok, Reddit, Patreon, Telegram), KPIs — grounded in a stored playbook, with build-failing transparency/brand-safety guardrails. Use on /simp-it or any AI influencer / virtual persona / brand-mascot / growth-persona request. Never adult, sexual, or suggestive.

YAMLskills/simp-it/agents/openai.yaml200B · 2026-07-23 · 2:27 PM

display_name: "Simp It

YAMLskills/simp-it/reach-playbook.yaml20.4KB · 2026-07-22 · 12:12 AM

reach-playbook.yaml — the knowledge core of /simp-it.

MDskills/simp-it/references/playbook.md11.2KB · 2026-07-21 · 11:46 AM

The Virtual-Influencer Reach Playbook (SFW / brand-safe) — grounding for /simp-it

MDskills/skin-it/SKILL.md9.5KB · 2026-07-22 · 5:15 PM

Suggest innovative, domain-matched, MOTION-FORWARD design skins for a site or app — palette, type, layout archetype, signature motif, named motion system — grounded in a stored library of real shipping templates + the 2026 animated-web canon. Proposes a safe-fit, a bold, and a wildcard skin, each with copy-paste :root tokens. Use on /skin-it or any design-direction, aesthetic, look-and-feel, animation/motion, "make it more engaging", or moodboard request.

YAMLskills/skin-it/agents/openai.yaml196B · 2026-07-23 · 2:27 PM

display_name: "Skin It

MDskills/skin-it/references/design-corrections-log.md3.1KB · 2026-07-18 · 5:18 PM

Design-corrections log — real skin corrections as skin-it grounding

MDskills/skin-it/references/envato-and-motion.md4.0KB · 2026-07-18 · 10:37 AM

This is the motion-grounding reference for skin-it, added after a dogfood finding that an early design pick was not engaging enough. It surveys Envato UI/UX templates and the 2026 animated-web canon, naming scroll sequences, parallax, kinetic type, micro-interactions, meaningful transitions, WebGL/3D, broken grids, 60fps discipline, and reduced-motion fallbacks.

e.g. It treats a scroll-triggered sequence where the scroll itself becomes the narrative timeline as a grounded motion pattern, not as vague advice to add animation.

UI — User InterfaceUX — User ExperienceWebGL — Web Graphics Library3D — Three-dimensional graphics60fps — Sixty rendered frames per secondCSS — Cascading Style SheetsEnvato — The referenced design-template marketplaceFigma — The referenced interface-design tool

MDskills/skin-it/references/lovable-gallery.md7.8KB · 2026-07-18 · 12:32 AM

This is the raw provenance corpus behind skin-library.yaml. It lists roughly 130 named Lovable templates across SaaS/apps, internal tools, ecommerce, portfolios, landing pages, blogs, music, events, services, product management, developer tools, resumes, and luxury, with the recurring aesthetic descriptors that the library distills.

e.g. The corpus names `Terminal Developer Portfolio` as a terminal/developer exemplar and `VOID Ecommerce` as a monochrome brutalist storefront exemplar.

SaaS — Software as a ServiceUI — User InterfaceUX — User ExperienceAI — Artificial IntelligenceCRM — Customer Relationship ManagementATS — Applicant Tracking SystemPDF — Portable Document FormatLovable — The template-gallery source named by this corpus

MDskills/skin-it/references/surface-patterns.md6.0KB · 2026-07-18 · 11:16 AM

This is the surface-level design reference for skin-it when the user names a page or component instead of a whole site. It records login and dashboard archetypes, hierarchy rules, component sets, motion patterns, accessibility needs, Dribbble screenshot provenance, and a wider pattern list for pricing, onboarding, checkout, settings, empty states, data tables, chat, and orchestration-relevant surfaces.

e.g. For B2B login, it makes email/SSO primary and social login secondary; for consumer login, social login is primary and email sits below.

UI — User InterfaceUX — User ExperienceB2B — Business to BusinessSSO — Single Sign-OnOAuth — An authorization protocol commonly used for social loginKPI — Key Performance IndicatorCSS — Cascading Style SheetsDribbble — The referenced design-gallery sourceAI — Artificial Intelligence⌘K — The keyboard shortcut convention for a command palettea11y — Accessibility shorthand

YAMLskills/skin-it/skin-library.yaml86.8KB · 2026-07-18 · 11:29 AM

This is the structured design corpus consumed by /skin-it. It defines aesthetic families with vibes, palettes, type pairings, layout archetypes, motion motifs and intensities, best-fit domains and buyers, avoid lists, and exemplars, plus reusable motion vocabulary, category rankings, wildcard moves, and surface patterns.

e.g. The `terminal-developer` family uses a near-black background, phosphor accent, monospace type, a console-window layout, and a typing-with-cursor motion motif.

YAML — The structured corpus formatWCAG-AA — Web Content Accessibility Guidelines, level AACSS — Cascading Style SheetsUI — User InterfaceUX — User Experience60fps — Sixty rendered frames per seconda11y — Accessibility shorthandLovable — The gallery source used to distill the familiesWebGL — Web Graphics Library

MDskills/slopless/SKILL.md6.3KB · 2026-08-07 · 2:20 PM

The anti-slop design GATE: critique a UI against the AI-slop uniform (purple/violet gradients, one generic thick sans, centered card-stacks, pillow buttons, soft-shadow-everything, blob backgrounds) and enforce craft — a planted stake, the generic cut, one deliberate human detail. Pairs with /skin-it (it PICKS the direction; slopless GUARDS the build). Use on /slopless, "does this look AI-generated / generic", or any pre-ship UI taste check.

YAMLskills/slopless/agents/openai.yaml192B · 2026-07-23 · 2:27 PM

display_name: "Slopless

MDskills/small-window/SKILL.md1.3KB · 2026-07-22 · 5:07 PM

Enter (or exit) the small-window operating mode — the tight-context discipline for sessions on a small window or fast-capping plan. Invoke with /small-window (or /small-window off to exit). God-kit privilege: in derived kits that pin the mode (e.g. entabeni) it is a REQUIREMENT and cannot be turned off.

YAMLskills/small-window/agents/openai.yaml188B · 2026-07-23 · 2:27 PM

display_name: "Small Window

MDskills/test-it/SKILL.md3.6KB · 2026-07-25 · 6:40 PM

Lay out the complete USER TESTING plan for a project or application — every user flow enumerated, each with numbered step-by-step instructions a human tester can follow cold, the expected result at every step, and a pass/fail record to fill in. Use on /test-it, "write the test plan", "how do I user-test this", after a build completes, or before handing anything to a human tester.

MDskills/verdict/SKILL.md1.9KB · 2026-07-18 · 4:32 PM

GOD-KIT ONLY. The claim-vs-canon lane — adjudicate "is this claim true per our canon?" through the NotebookLM notebook (grounded in the kit's own sources), never an executor's own say-so. Use when the user says /verdict, when two sources conflict, or when a worker's claim needs independent adjudication.

YAMLskills/verdict/agents/openai.yaml205B · 2026-07-23 · 2:27 PM

display_name: "Verdict (god kit)

MDskills/watch-local/SKILL.md6.0KB · 2026-08-06 · 9:02 PM

Arm the observer pattern on the FREE local model — 6 session watchers feeding the local gpt-oss router, fully detached and idempotent, $0/idle, no long-lived Claude window. Use on /watch-local, "start the local watch/observer", "arm the free watcher". The cheap counterpart to /continue observer.

YAMLskills/watch-local/agents/openai.yaml182B · 2026-07-23 · 2:27 PM

display_name: "Watch Local

PYskills/watch-local/payload/agy-all-watch.py12.4KB · 2026-07-28 · 9:02 PM

Watch the Antigravity CLI (agy); emit new human messages and error lines for the observer

PYskills/watch-local/payload/claude-all-watch.py6.6KB · 2026-08-08 · 3:12 AM

GENERATED FROM tools/claude-all-watch.py — DO NOT EDIT (regenerate: python3 tools/generate-watch-payload.py)

PYskills/watch-local/payload/codex-all-watch.py8.3KB · 2026-08-08 · 3:12 AM

GENERATED FROM tools/codex-all-watch.py — DO NOT EDIT (regenerate: python3 tools/generate-watch-payload.py)

PYskills/watch-local/payload/grok-all-watch.py14.8KB · 2026-07-23 · 10:30 AM

Watch grok CLI sessions; emit new human messages and error lines for the observer router.

FILEskills/watch-local/payload/l123-neg.txt127B · 2026-07-22 · 1:05 AM

as expected

FILEskills/watch-local/payload/l123-pos.txt398B · 2026-07-22 · 8:52 PM

cap[- ]?death

PYskills/watch-local/payload/observer-router.py87.7KB · 2026-08-06 · 9:02 PM

Local, free, always-on observer/router for Wolf's agent fleet.

MDskills/watch-local/payload/observer-rubric.md7.8KB · 2026-07-23 · 10:33 AM

Observer Routing Rubric

PYskills/watch-local/payload/observify-inbox-watch.py2.4KB · 2026-08-08 · 3:12 AM

GENERATED FROM tools/observify-inbox-watch.py — DO NOT EDIT (regenerate: python3 tools/generate-watch-payload.py)

SHskills/watch-local/payload/remote-loop.sh867B · 2026-07-23 · 10:44 AM

remote-loop.sh HOST REMOTE_CMD LABEL — one self-healing SSH watcher loop for a remote machine.

PYskills/watch-local/payload/verify-payload.py19.4KB · 2026-07-23 · 12:03 PM

Regression gate for the watch-local payload (L154-L157 + L129/L155 counter/burst/wake/feed-health behaviors) -- run: python3 verify-payload.py

SHskills/watch-local/payload/watch-local.sh5.6KB · 2026-07-31 · 9:40 AM

watch-local.sh — arm the LOCAL-MODEL observer pattern: 6 session watchers feeding the free

Tools

67 files — fleet apparatus copied into _build/

PYtools/article1-pretool-guard.py12.5KB · 2026-08-08 · 2:53 AM

Claude Code PreToolUse enforcement for Constitution Article I.

SHtools/assess.sh4.2KB · 2026-07-12 · 2:04 PM

This is the read-only fleet state classifier used by /continue. It combines progress, done markers, process liveness, unit ledgers, open CONTROL directives, ground-truth atom totals, and blockers to classify each unit as RUNNING, DEAD, PAUSED, STOPPED, BLOCKED, DONE-CLAIMED, DONE-MARKED, or a directive-conflicted state with a suggested action.

e.g. It checks liveness before pending directives so an alive worker is reported RUNNING rather than incorrectly suggested for relaunch.

sh — Shell script file suffixCONTROL — Per-unit command busSTATUS — Per-unit ledgerDONE — Completion claim or markerL18 — The lesson fixing classifier ordering and scope measurementF8 — Ground-truth substitution/configuration fixF13 — Single-worker lease guard identifierGit — Distributed version-control systemPID — Process identifier

SHtools/audit-done.sh5.9KB · 2026-07-12 · 2:04 PM

This is the integrity gate for a unit's DONE claim. It compares ledger atom claims with ground-truth counts and spec totals, checks substantive completion files, requires a structural PRODUCT WALK section with per-flow PASS/FAIL evidence, and exits nonzero when the claim is incomplete, fabricated, or un-auditable.

e.g. A line such as `F1 — PASS (evidence: screenshot.png)` satisfies the product-walk shape, while a bare word like `dogfood` does not.

sh — Shell script file suffixDONE — Completion claim or markerSTATUS — Per-unit ledgerDOGFOOD — Using the product as a real userPRODUCT WALK — Cold-user flow verificationPASS/FAIL — The required per-flow verdict formATOM-ID — Countable work-item identifierF7 — The product-walk format synchronization fixF8 — The unit/slug ground-truth substitution fixL28 — The product-walk integrity lessonYAML — The control-bus/config format

PYtools/bookit/__init__.py74B · 2026-08-07 · 2:00 PM

book-it: curate a read-only reMarkable library into EPUB editions.

PYtools/bookit/board.py10.4KB · 2026-08-07 · 2:00 PM

Observify kanban board lens for book-it.

PYtools/bookit/bookit.py14.9KB · 2026-08-08 · 12:20 AM

Command-line entry point for book-it.

PYtools/bookit/catalog.py6.2KB · 2026-08-07 · 2:00 PM

Read-only cataloguing of the reMarkable Desktop content store.

PYtools/bookit/compile.py50.9KB · 2026-08-07 · 7:33 PM

EPUB3 compiler for an approved book-it manifest.

PYtools/bookit/consumer_render.py24.4KB · 2026-08-07 · 8:35 PM

G4 — the CONSUMER-SIMULATION RENDER gate (L202, L201).

PYtools/bookit/delete_verify.py11.4KB · 2026-08-07 · 8:20 PM

delete_verify.py — L205: a delete at one replica is a change the other replica will undo.

PYtools/bookit/deliver.py4.7KB · 2026-08-08 · 2:17 AM

Primary, direct-to-device delivery through rmapi.

PYtools/bookit/design/__init__.py839B · 2026-08-07 · 7:27 PM

The OPERATION print design system — the device format book-it renders into.

PYtools/bookit/design/geometry.py11.1KB · 2026-08-07 · 7:59 PM

Device geometry, derived — not asserted.

FILEtools/bookit/design/operation.sty36.9KB · 2026-08-07 · 8:24 PM

% =====================================================================

PYtools/bookit/design/palette.py7.4KB · 2026-08-07 · 7:25 PM

The edition's colour system, and the Pygments theme built on it.

PYtools/bookit/ledger.py9.7KB · 2026-08-08 · 12:15 AM

Edition ledger and deliberately conservative curation proposals.

PYtools/bookit/lenses.py11.8KB · 2026-08-08 · 12:20 AM

Selection lenses and manifest construction for book-it.

PYtools/bookit/pi.py3.0KB · 2026-08-07 · 7:37 PM

Opt-in Pi backup and Google Drive delivery lane.

PYtools/bookit/render_pdf.py64.5KB · 2026-08-08 · 12:18 AM

The PRIMARY renderer: an OPERATION field manual PDF, sized to the panel 1:1.

PYtools/bookit/tests/test_delivery_naming.py3.2KB · 2026-08-08 · 2:19 AM

from pathlib import Path

PYtools/bookit/tests/test_ledger_loading.py1.6KB · 2026-08-08 · 12:44 AM

from pathlib import Path

PYtools/bookit/tests/test_medium_prior_flags.py3.0KB · 2026-08-08 · 12:45 AM

from collections import Counter

PYtools/bookit/tests/test_medium_self_ingestion.py2.1KB · 2026-08-08 · 12:44 AM

from collections import Counter

PYtools/bookit/tests/test_pdf_missing_glyphs.py3.9KB · 2026-08-08 · 12:45 AM

from pathlib import Path

PYtools/bookit/tests/test_pdf_quarantine.py3.4KB · 2026-08-08 · 12:45 AM

from pathlib import Path

PYtools/build-index.py10.1KB · 2026-08-07 · 8:26 PM

build-index.py — regenerate kernel/INDEX.tsv, the kit's card catalog.

SHtools/bus-sweep.sh824B · 2026-07-06 · 9:15 AM

This is the fleet-wide control-bus parser check used after unattended windows or injector repairs. It requires Python and PyYAML before judging parse state, loads every unit CONTROL.yaml, prints corrupt paths, and exits nonzero if any bus fails to parse.

e.g. Its clean result is `bus sweep clean (N buses parse)`; a malformed unit bus prints `CORRUPT: <path>`.

sh — Shell script file suffixYAML — The control-bus data formatPyYAML — Python's YAML parser libraryCONTROL — Per-unit directive busL22 — The bus-can-be-the-fault lessonL7 — The prove-the-bus lessonL — Lesson/scar identifier

PYtools/check-canon-reads.py6.6KB · 2026-07-29 · 7:54 AM

Usage: python3 tools/check-canon-reads.py <session-transcript.jsonl>

SHtools/check-generated.sh3.9KB · 2026-07-29 · 7:34 AM

Usage: bash tools/check-generated.sh

PYtools/check-install-lane.py10.3KB · 2026-07-29 · 7:46 AM

Usage: python3 tools/check-install-lane.py <session-transcript.jsonl>

PYtools/check-workflow-args.py13.7KB · 2026-08-08 · 12:17 AM

check-workflow-args.py — G-ARGS-COERCED (L208): `args` crosses the Workflow boundary as a

PYtools/claude-all-watch.py6.5KB · 2026-07-16 · 3:01 PM

This is the Claude Code transcript observer used by /learn. It adopts existing JSONL sessions at EOF, follows new sessions from byte zero, skips its own session with a required --self ID, filters markup and re-render noise, and emits human messages plus declaration-shaped outcomes such as gates, exits, caps, shipping, markers, and audits.

e.g. An assistant line containing `GATE FAIL` becomes an OUTCOME event, while a bare prose word `failed` is ignored to avoid monitor floods.

py — Python source file suffixJSONL — JSON Lines transcript formatJSON — JavaScript Object NotationAPIERROR — An emitted application-programming-interface error eventEXIT_CODE — Process exit status markerL8 — Monitor-flood lessonL17 — Domain-vocabulary false-positive lessonL41 — Adopt-existing-session-at-EOF lessonL91 — Harvest-the-whole-loop lessonL109 — Context-blocked session lessonPTY — Pseudo-terminal

PYtools/codex-all-watch.py8.2KB · 2026-07-28 · 9:07 AM

This is the Codex-session edition of the /learn observer. It reads rollout JSONL files, distinguishes interactive sessions from codex_exec workers, follows human and final agent messages for interactive sessions, emits worker failures and caps without a reasoning firehose, and adopts existing rollouts at EOF.

e.g. A worker quota event is emitted as `[codex <sid8>] CAP usage-limit: ...`, while routine worker task_complete events are suppressed.

py — Python source file suffixJSONL — JSON Lines transcript formatJSON — JavaScript Object NotationCAP — Usage-cap event labelcwd — Current working directoryL8 — Monitor-noise lessonL41 — Adopt-existing-session-at-EOF lessonL90 — Multi-runtime observer-scope lessonL91 — Harvest-the-whole-loop lesson429 — HTTP rate-limit statusCodex — The external coding-agent runtime

SHtools/dashboard.sh1.4KB · 2026-07-03 · 2:06 PM

This is the human-facing one-shot live dashboard for workers. It prints one block per queued unit with an alive marker, current atom, last gate, blockers, and the last meaningful narration line after stripping diffs and execution noise; it is meant for repeated `watch` display, not raw log tailing.

e.g. A row looks like `● unit | current-atom | last-gate | blockers`, followed by a single concise narration line.

sh — Shell script file suffixSTATUS — Per-unit ledgerCT1 — The delegate-reading context ruleGit — Distributed version-control systemPID — Process identifier

SHtools/dispatch.sh11.2KB · 2026-08-08 · 12:38 AM

This is the detached multi-vendor side-task courier. It requires a packet, enforces one live vendor run per task, starts codex/claude/agy/grok/ollama with their headless postures, records raw output and exit state, provides bounded wait and nonblocking status, and leaves normalization to normalize-report.sh.

e.g. `dispatch.sh wait AUD-003` returns `RUNNING` with exit 10 after its bounded chunk, so the caller can poll again without exceeding the harness shell cap.

sh — Shell script file suffixCLI — Command-Line InterfacePTY — Pseudo-terminalJSON — JavaScript Object NotationSSH — Secure ShellKB — Knowledge-base pattern identifierL5 — Usage-cap lessonL116 — Turnkey delegation lessonAPI — Application Programming InterfaceREPORT_SCHEMA — Path to the side-task report schemaVENDOR_WAIT_S — Bounded wait interval settingVENDOR_MAX_S — Detached-run deadline setting

FILEtools/dispatch.sh.bak-pre-eng-increment1-20260808T00491111.2KB · 2026-08-08 · 12:38 AM

Multi-vendor side-task dispatch (executors.yaml `layers` + `routing`).

SHtools/dogfood-verify.sh3.4KB · 2026-07-12 · 2:04 PM

This is the machine-runnable half of a unit dogfood check. It runs the unit's own tests, seed twice for idempotency, the same gate reviewers use, service boot and health probes when configured, and a comparison of spec flow IDs with DOGFOOD.md coverage; human walkthroughs remain separate.

e.g. If a package declares a `seed` script, it runs it twice and reports `seed x2 (idempotent) PASS` only when the second run also succeeds.

sh — Shell script file suffixDOGFOOD — Real use of the built artifactAPI — Application Programming InterfaceHTTP — HyperText Transfer ProtocolL15 — Verify through the unit's own command lessonL16 — Dogfood is run, not merely documentedF — User-flow identifierJSON — JavaScript Object NotationPID — Process identifier

PYtools/fanout-census.py11.0KB · 2026-08-07 · 8:21 PM

fanout-census.py — L203: when N workers fill the same field, that field's SHAPE is a spec.

FILEtools/fixtures/workflow-args/BAD-wf_b52ac5fc-8fa.js4.1KB · 2026-08-08 · 12:15 AM

export const meta = {

FILEtools/fixtures/workflow-args/BAD-wf_bed63bba-46d.js9.5KB · 2026-08-08 · 12:15 AM

export const meta = {

FILEtools/fixtures/workflow-args/GOOD-wf_59d5074b-1c9.js9.5KB · 2026-08-08 · 12:15 AM

export const meta = {

FILEtools/fixtures/workflow-args/GOOD-wf_823efbee-b73.js3.9KB · 2026-08-08 · 12:16 AM

export const meta = {

FILEtools/fixtures/workflow-args/GOOD-wf_86f560d8-b3f.js4.7KB · 2026-08-08 · 12:16 AM

export const meta = {

FILEtools/fixtures/workflow-args/GOOD-wf_9258c732-a8f.js9.4KB · 2026-08-08 · 12:15 AM

export const meta = {

PYtools/frontmatter.py16.0KB · 2026-08-07 · 8:20 PM

frontmatter.py — the kit's ONE loader for `---`-delimited YAML front matter.

SHtools/gate.sh3.0KB · 2026-08-08 · 12:18 AM

This is the per-unit merge gate and the single definition of green. It loads project.env, runs the configured project gate or package gate, refuses an empty gate as a lie, prints each check, and exits nonzero with GATE FAIL when any check fails.

e.g. With no GATE_CMD, no package `gate` script, and no uncommented checks, it emits `GATE FAIL: no checks defined` instead of passing vacuously.

sh — Shell script file suffixGATE_CMD — Configured command used as the unit gatePASS/FAIL — Gate result statesCI — Continuous IntegrationL15 — The verify-via-own-command lessonWF-4 — The empty-gate failure classGit — Distributed version-control system

PYtools/gen-surfaces.py12.3KB · 2026-07-28 · 8:25 PM

Render the constitution and its role-scoped instruction surfaces from YAML.

PYtools/generate-watch-payload.py2.9KB · 2026-07-31 · 3:40 AM

generate-watch-payload.py

SHtools/inject.sh3.0KB · 2026-07-06 · 9:15 AM

This is the append-only control-bus writer. It validates directive type and argument shape, requires a YAML parser, normalizes inline directives, guarantees a trailing newline, emits multiline instructions as a block scalar, parse-checks the whole file, and rolls back a poisoning append.

e.g. `inject.sh <unit> pause general "stop after checkpoint"` appends a numbered OPEN directive and reports `parse-verified`.

sh — Shell script file suffixYAML — The control-bus data formatPyYAML — Python's YAML parser libraryOPEN — Directive status awaiting execution and acknowledgementACK — Worker acknowledgementL7 — The bus liveness lessonL22 — The corrupt-bus lessonKB — Knowledge-base pattern identifierISO8601 — International date/time format

PYtools/install-hooks.py9.4KB · 2026-07-31 · 10:17 AM

install-hooks.py — keep the fleet's project-local SessionStart hook config current (hooks-local-0731).

PYtools/install-surfaces.py8.9KB · 2026-07-31 · 9:49 AM

install-surfaces.py — install refreshed constitution surfaces into fleet workspaces.

SHtools/keep-fed.sh3.9KB · 2026-07-15 · 2:07 PM

This is the supervisor loop that repeatedly relaunches one-shot workers until audited done markers cover all requested slugs or a safe stop condition occurs. It audits fresh markers, compares Git HEAD fingerprints for progress, stops on pool failure, no-progress rounds, maximum rounds, or a competing supervisor lock.

e.g. After three rounds with no Git HEAD movement it exits 4 as a silent-dead pool instead of churning fifteen retries.

sh — Shell script file suffixGit — Distributed version-control systemHEAD — The current Git commit referenceDONE — Completion marker/claimMAX_ROUNDS — Maximum relaunch-round settingNO_PROGRESS_MAX — Allowed consecutive no-progress roundsAUDIT_MARKERS — Setting that gates markers through audit-done.shL4 — Resume-by-ledger lessonL24 — Marker-is-a-claim lessonL98 — Single supervisor lessonPID — Process identifier

PYtools/kit-codex-session-start.py6.2KB · 2026-07-28 · 4:18 PM

kit_session_start.py — Codex CLI SessionStart hook: a Codex session that has a kit STARTS

SHtools/kit-remember-hook.sh8.6KB · 2026-07-31 · 10:16 AM

kit-remember.sh — SessionStart hook: a session that has a kit STARTS with the canon (L161).

SHtools/lib.sh3.1KB · 2026-07-28 · 6:11 PM

This is the shared shell helper library for recurring structural fixes. It provides safe numeric grep counts, trailing-newline repair, and a distinct PyYAML/python3 dependency preflight so injectors and bus sweeps do not confuse a missing parser with a corrupt bus.

e.g. `require_yaml_parser` exits with `MISSING DEPENDENCY: PyYAML` before any parse verdict when the yaml module is absent.

sh — Shell script file suffixPyYAML — Python's YAML parser libraryYAML — The control-bus/config data formatL7 — The prove-the-bus lessonL18 — The prose-lesson-is-not-a-fix lessonpython3 — Python 3 interpreter

SHtools/normalize-report.sh4.2KB · 2026-07-23 · 10:29 AM

This is the report normalizer for vendor side-task runs. It unwraps Grok envelopes, accepts deterministic Codex last-message or raw JSON first, uses brace-aware extraction, optionally asks free local Ollama glue to coerce a transcript, validates required fields with jq, and treats invalid normalization as a failed run.

e.g. A valid normalized report must contain string task_id/agent/model/workdir, a status in `done|blocked|failed|partial`, and arrays for files_touched, commands_run, evidence, and uncertainties.

sh — Shell script file suffixJSON — JavaScript Object NotationGrok — The xAI side-task executorCodex — The OpenAI coding-agent executorOllama — The local model runnerjq — Command-line JSON processorSCHEMA — The report validation contractAPI — Application Programming InterfaceKB — Knowledge-base pattern identifier

PYtools/observify-inbox-watch.py2.3KB · 2026-07-17 · 8:24 PM

This is the opt-in receiver that closes the loop between the Observify dashboard and a running session. It tails the current session's JSONL inbox, emits queued and newly written interjections as task notifications, and intentionally watches only the session ID passed at startup.

e.g. Running `python3 observify-inbox-watch.py --session <this-session-id>` prints `[INTERJECTION → <short>] <text>` for each directive.

py — Python source file suffixJSONL — JSON Lines inbox formatJSON — JavaScript Object NotationAPI — Application Programming InterfaceObservify — The live-session dashboard named by this tool

SHtools/offload.sh6.2KB · 2026-07-12 · 1:59 PM

This is the SSH machine-fallback tool for moving compute when the current host is under memory pressure. It measures free memory on macOS or Linux, bootstraps a remote kit and adapted environment with rsync, verifies remote git and executor access, launches a detached remote pool, and preserves the distinction that offload moves compute, not provider quota.

e.g. `offload.sh auto 3 a b c` uses the local pool when free memory is healthy, otherwise selects the first reachable OFFLOAD_HOST and starts a remote pool.

sh — Shell script file suffixSSH — Secure Shellrsync — Remote file synchronization utilityRAM — Random-Access MemoryOS — Operating SystemGit — Distributed version-control systemPID — Process identifierL4 — Resume-by-ledger lessonL7 — Prove-the-channel lessonREMOTE_KIT_DIR — Remote tool directory settingREMOTE_UNITS_ROOT — Remote workspace root settingREMOTE_CLONE_BASE — Remote Git clone base setting

SHtools/packet-lint.sh3.1KB · 2026-07-28 · 8:52 PM

packet-lint.sh — Constitution Article VIII: the packet is the contract.

SHtools/pool.sh7.4KB · 2026-07-14 · 12:06 PM

This is the concurrent one-shot worker launcher. It reads project.env, probes the actual executor pool, resolves per-slug executors and models, renders worker-prompt.md, refuses unresolved placeholders or missing workspaces, enforces a per-unit lease, detaches the worker from the caller's process group, logs the real exit code, and waits for up to K workers.

e.g. Before launching it checks that `worker-prompt.md` is beside pool.sh and rejects a residual `{UNIT_VAR}` placeholder rather than spending a worker turn guessing.

sh — Shell script file suffixK — Maximum number of concurrent workersCLI — Command-Line InterfaceGit — Distributed version-control systemPID — Process identifierL29 — Pre-launch quota-probe lessonL45 — Worker cwd/grounding lessonL63 — One-shot worker lessonL66 — Detached-session/process-group lessonF6 — Worker prompt placeholder substitution fixF13 — Per-unit lease fixEXIT_CODE — Recorded process exit status

SHtools/progress.sh1.8KB · 2026-07-06 · 1:21 PM

This is the fleet progress calculator. For each queued slug it substitutes unit and slug paths into configured total/done commands, counts ground-truth atoms rather than ledgers, loudly excludes a misconfigured zero total, floors the denominator when scope grows, and prints one percentage line suitable for change-triggered monitoring.

e.g. If a unit grows from 10 planned atoms to 12 completed atoms, it reports `12/12+` rather than a misleading percentage above 100.

sh — Shell script file suffixATOM_NAME — Configured name for a progress atomATOMS_TOTAL_CMD — Command returning a unit's spec totalATOMS_DONE_CMD — Command returning ground-truth completed atomsF8b — The loud zero-total configuration guardL18 — The scope-measurement lessonGit — Distributed version-control system

PYtools/remember.py11.1KB · 2026-07-25 · 3:23 AM

remember.py — the session-start canon read, made affordable and COMPLETE.

SHtools/tools-gate.sh4.4KB · 2026-07-28 · 6:29 PM

tools-gate.sh — Constitution Article VII: headless proof & first output.

PYtools/transcript_adapter.py7.8KB · 2026-07-29 · 7:34 AM

Normalize supported Claude Code and Codex JSONL transcripts for kit checks.

SHtools/verdict.sh2.0KB · 2026-07-10 · 7:01 PM

This is the shell wrapper for the NotebookLM layer-3 verdict oracle. It checks for notebooklm and jq, asks the notebook to grade a claim strictly against loaded sources with citations, extracts the last parseable JSON object, defaults source-less responses to unverified, and emits a JSON result for human finalization.

e.g. The request explicitly says: if no source speaks to the claim, the verdict MUST be `unverified`.

sh — Shell script file suffixJSON — JavaScript Object Notationjq — Command-line JSON processorCLI — Command-Line InterfaceAPI — Application Programming InterfaceL — Lesson/scar identifierKB — Knowledge-base pattern identifiervalidated — Citation supports the claimunverified — No source speakscontradicted — Source conflicts and blocks

PYtools/verify-conductor.py6.0KB · 2026-07-28 · 8:52 PM

verify-conductor.py — Constitution Article I enforcement on a session transcript.

SHtools/watch.sh763B · 2026-07-03 · 3:38 PM

This is the cheapest cross-unit survey tool. It prints one block per unit with the count of open control directives and the compact STATUS fields for milestone, current atom/ticket, last gate, and blockers, making it suitable for a heartbeat instead of raw log reading.

e.g. A unit block begins `=== <unit> OPEN-directives:<n>` and then shows its current ledger header.

sh — Shell script file suffixSTATUS — Per-unit ledgerCONTROL — Per-unit directive busCT1 — The delegate-reading context ruleGit — Distributed version-control system

Templates

16 files — skeletons instantiated per project

YAMLtemplates/CONTROL.yaml973B · 2026-07-03 · 1:25 PM

This is the per-unit YAML seed for the append-only reviewer command bus. It defines directive fields, the seven allowed types, worker acknowledgement expectations, and the open directive shape that inject.sh appends and workers consume at checkpoints.

e.g. A `correct` directive targets an ATOM-ID, names an optional guardrail, carries an instruction, and stays `status: OPEN` until the worker ACKs it.

YAML — The control-bus data formatACK — Acknowledgement that a worker executed a directiveATOM-ID — Identifier for a countable work atomG-ID — Build-failing guardrail identifierISO8601 — International date/time representationL7 — The control-bus proof lesson

MDtemplates/STATUS.md628B · 2026-07-03 · 1:25 PM

This is the per-unit Markdown ledger template. Workers overwrite its small header with the current unit, milestone, atom, gate, landed evidence, promotions, and blocker state, then append notes about assumptions, directives, self-dogfood, and anything a reviewer needs to trust.

e.g. The header includes `last-gate: PASS|FAIL @ <ISO8601>` and `blockers: none|PAUSED|STOPPED|DONE`.

MD — Markdown document formatISO8601 — International date/time representationG-STATUS-CURRENT — Guardrail requiring the ledger header to reflect realityDONE — Completion claim/markerPAUSED — Worker parked for later resumeSTOPPED — Worker stopped by directive

MDtemplates/authorized-context-header.md3.7KB · 2026-07-17 · 3:13 PM

This is the honest context-framing template for authorized security work or a persistent authored persona. It supplies true facts that were missing, explicitly forbids using the header as a jailbreak or guardrail bypass, routes to a policy-fit lane first, and stops if the register remains outside every available lane.

e.g. The security variant asks for the authorization basis, in-scope assets, out-of-scope assets, defensive goal, and disclosure path before a session starts.

MD — Markdown document formatL109 — The context-blocked-session lessonL112 — The authorized dual-use routing/framing lessonTCC — macOS Transparency, Consent, and ControlURL — Uniform Resource LocatorAPI — Application Programming Interface

MDtemplates/codex-agents-block.md2.5KB · 2026-07-28 · 4:18 PM

SHtemplates/ctx-check.sh1.7KB · 2026-07-10 · 9:01 PM

This is the agent-side context meter for CT10. It reads the latest Claude transcript usage record, adds input, cache-read, and cache-creation tokens, compares them with CTX_WINDOW, and prints the physical window occupancy that should drive pause and handoff decisions.

e.g. It reports a value like `~47% used (470,000 / 1,000,000 tokens)` from the latest usage record rather than trusting a harness percentage that may omit overhead.

CT10 — The know-your-meter context-thrift ruleJSONL — JSON Lines transcript formatCTX_WINDOW — Configured context-window size in tokensJSON — JavaScript Object NotationClaude — The model/runtime whose transcript is readRAM — Random-Access Memory

YAMLtemplates/guardrails.yaml2.5KB · 2026-08-08 · 12:19 AM

This is the project guardrail skeleton filled from intake. It names invariants such as no-touch-existing, standalone units, a nonempty gate, control-bus ordering, current status, and use-before-done, and pairs each with an enforcement mechanism or a placeholder for a project-specific failing check.

e.g. G-USE-BEFORE-DONE requires realistic self-use and a cold user completing every core flow, and points the reviewer to done-audit evidence.

YAML — The guardrail configuration formatG- — Build-failing guardrail identifierG-NO-TOUCH-EXISTING — Guardrail limiting a worker to its own workspaceG-GATE — Guardrail making gate.sh the definition of greenG-STATUS-CURRENT — Guardrail requiring evidence-backed ledger stateG-USE-BEFORE-DONE — Guardrail requiring self-use and cold-user completion

MDtemplates/handoff-packet.md3.2KB · 2026-07-29 · 7:18 AM

This is the packet template for a bounded vendor side-task. The orchestrator fills it before dispatch with task identity, objective, workdir, scope boundaries, context, an exact JSON evidence format, verification commands, stop conditions, and—on escalation—the prior report verbatim.

e.g. It says `No packet, no dispatch` and requires the final message to be one JSON object matching report.schema.json with task_id, agent, model, status, workdir, files_touched, commands_run, evidence, and uncertainties.

MD — Markdown document formatJSON — JavaScript Object NotationTASK-ID — Side-task identifierWORKDIR — The worker's allowed working directoryIN-SCOPE — Allowed files or surfacesOUT-OF-SCOPE — Explicitly forbidden files or surfacesKB — Knowledge-base pattern identifierPRIOR-ATTEMPT — Verbatim previous report used during escalation

MDtemplates/instance-rollout.md3.3KB · 2026-07-31 · 10:20 AM

God-kit instance rollout — the five parts (L188)

YAMLtemplates/pipeline.yaml3.8KB · 2026-07-06 · 1:20 PM

This is the canonical per-project pipeline template. It defines paths, atoms and evidence, the worker loop, milestones, gate command, control-bus obligations, completion requirements including seed/self-use/product walk/runbook, and the integrity rule that an evidence-less ledger claim voids the build.

e.g. Its work loop says a worker must re-read CONTROL.yaml, execute and ACK open directives, implement one atom, re-read before a checkpoint, run the gate, and land evidence only after PASS.

YAML — The machine-readable pipeline formatGit — Distributed version-control systemATOM-ID — Countable work-item identifierACK — Directive acknowledgementDOGFOOD — Using the built artifact as a real userPRODUCT WALK — Cold-user completion of every core flowG-ID — Guardrail identifierSTATUS — Per-unit ledger fileCONTROL — Per-unit directive-bus file

ENVtemplates/project.env8.0KB · 2026-07-15 · 11:29 PM

This is the environment template that parameterizes every fleet tool. It holds project paths, executor and model settings, quota probes, atom ground-truth commands, per-unit ledger names, vendor side-task options, SSH offload settings, and the done-audit contract.

e.g. It defines `ATOMS_DONE_CMD` as a command that must count real commits/files rather than ledger claims and warns that `{unit}` and `{slug}` are literal substitution tokens.

ENV — Environment-variable configurationYAML — The format used by pipeline and control filesSSH — Secure ShellCLI — Command-Line InterfaceAPI — Application Programming InterfaceJSON — JavaScript Object NotationG-ID — Guardrail identifierATOM — Countable work itemOS — Operating SystemL5 — Usage-cap lessonL29 — Pre-launch quota-probe lessonF12 — Worker OS-sandbox guard identifierKB — Knowledge-base pattern identifier

JSONtemplates/report.schema.json1.3KB · 2026-07-10 · 7:01 PM

This is the JSON Schema for the single report object returned by every dispatched side-task. It requires task, agent, model, status, workdir, touched files, commands with exit codes, evidence, and uncertainties, allows a nullable stop condition, and forbids extra properties.

e.g. The status enum is exactly `done`, `blocked`, `failed`, or `partial`, and each commands_run item must contain `cmd` and integer `exit`.

JSON — JavaScript Object NotationSCHEMA — A formal structure and validation contractHTTP — HyperText Transfer ProtocolAPI — Application Programming InterfaceKB — Knowledge-base pattern identifier

SHtemplates/sandboxed-claude.sh693B · 2026-07-08 · 5:30 AM

This is the launcher wrapper that runs a Claude worker under the rendered macOS Seatbelt profile. It disables auto-updates and delegates the actual boundary to worker-sandbox.sb, confining writes and denying sensitive UI, TCC, settings, sudo, and remote-access paths.

e.g. It executes the sandbox-exec wrapper with worker-sandbox.sb and the configured Claude binary.

OS — Operating SystemTCC — macOS Transparency, Consent, and ControlSSH — Secure ShellCLI — Command-Line InterfaceSBPL — Sandbox Profile Language used by Seatbelt

SHtemplates/statusline-context.sh3.2KB · 2026-07-10 · 9:01 PM

This is a Claude Code statusline renderer for human-visible context occupancy. It reads JSON from stdin, uses native percentages when available, estimates from raw token counts when necessary, colors a ten-segment bar by threshold, honors NO_COLOR, and falls back to the model name when no context data exists.

e.g. It renders a line such as `Claude · ctx ▓▓▓▓▓░░░░░ 47% used (53% free)`.

JSON — JavaScript Object NotationNO_COLOR — Environment convention requesting no ANSI colorANSI — American National Standards Institute terminal escape-code conventionCLI — Command-Line Interfacectx — Context-window usage shorthand

MDtemplates/vendor-agents-block.md5.4KB · 2026-07-29 · 7:54 AM

<!-- >>> think-like-fable vendor bridge — managed by ship-kit.sh (not yet wired, see GROK.md/AGY.md TASK-3 note); edit canonical: think-like-fable/templates/vendor-agents-block.md

MDtemplates/worker-prompt.md6.7KB · 2026-07-22 · 4:54 PM

This is the agent-facing worker contract rendered by pool.sh for each unit. It binds the worker to the canonical spec, human register, one workspace and branch, one-shot foreground verification, control-bus checkpoints, evidence-backed atoms and milestones, cold-start product walks, shared-machine limits, and a parseable-bus stop rule.

e.g. It explicitly says the worker must not write DONE_FILE; the orchestrator mints the shared marker only after done-audit passes.

MD — Markdown document formatYAML — The structured prompt format after its title line is strippedACK — Directive acknowledgementVCS — Version Control SystemGit — Distributed version-control systemBC — Behavioral contractKB — Knowledge-base patternG-NO-TOUCH-EXISTING — Own-workspace boundary guardrailL39 — Reviewer-minted markers lessonL53 — One-shot worker cannot wait lessonL63 — No blind windows lessonG- — Guardrail identifier

SBtemplates/worker-sandbox.sb4.2KB · 2026-07-11 · 2:18 PM

This is the macOS Seatbelt policy rendered for a worker workspace. It denies file writes by default, allows the unit workspace and selected caches, makes most of the home directory unreadable, and hard-denies screen capture, UI control, settings changes, sudo, launch control, and SSH/scp/sftp.

e.g. The profile allows writes under `{WORKSPACE}` and `/private/tmp` but denies process execution of `/usr/bin/osascript`, `/usr/bin/sudo`, and `/usr/bin/ssh`.

SBPL — Sandbox Profile LanguageOS — Operating SystemTCC — macOS Transparency, Consent, and ControlSSH — Secure ShellUI — User InterfaceCLI — Command-Line InterfaceL37 — The browser/headless sandbox lessonL33 — The shared live desktop window lesson

Root docs

16 files — narrative canon + NotebookLM sources

MDAGY.md2.5KB · 2026-07-29 · 7:40 AM

AGY.md — running the kit under Google agy CLI (Antigravity)

MDCODEX.md7.6KB · 2026-08-06 · 11:49 PM

CODEX.md — running the kit's skills under Codex CLI

MDCONSTITUTION.md15.2KB · 2026-08-08 · 3:12 AM

MDCONTEXT_THRIFT.md5.1KB · 2026-07-22 · 4:54 PM

This is the human-readable rulebook for protecting the orchestrator's context window, which the kit treats as its scarcest resource. It prescribes delegated reading, filtered event monitors, cheap surveys before deep audits, model tiering, disk-backed state, batched human questions, and one-writer file discipline.

e.g. It says a noisy monitor should be stopped and re-armed with tighter filters immediately, rather than allowed to consume a turn per event.

CT — Context-thrift rule familyCT1-CT10 — The numbered context-thrift rulesL — Lesson or scar identifierRESUME.md — A disk-backed handoff/state snapshot

MDFEATURE_MAP_BUILD_SPEC.md14.5KB · 2026-07-29 · 8:21 AM

This is the strict production specification for a self-contained, multipage HTML feature map of the kit. It fixes the page order, semantic HTML, print and screen CSS, required lifecycle/orchestra/tool/template/integrity content, pinned spellings, and a QA checklist that prevents invented or garbled facts.

e.g. Its v2 addendum requires Page 7 to show all three verdicts: validated / unverified / contradicted, with contradicted blocking the decision.

HTML — HyperText Markup LanguageCSS — Cascading Style SheetsTOC — Table of ContentsQA — Quality assuranceWCAG-AA — Web Content Accessibility Guidelines, level AAP1-P12 — The twelve principle identifiersL1-L29 — Lesson/scar identifiers in the pinned terminologyCT1-CT10 — Context-thrift rule identifiersBC-* — Behavioral-contract evaluation casesKB-* — Knowledge-base pattern identifiersSSH — Secure Shell

MDGROK.md2.3KB · 2026-07-29 · 7:40 AM

GROK.md — running the kit under xAI Grok Build (grok-cli)

MDHOW_TO_USE.md13.4KB · 2026-07-29 · 8:23 AM

This is the command-by-command operating manual for installing, configuring, running, steering, auditing, and recovering a kit-managed fleet. It explains prerequisites, the project.env surface, the control bus, quota and SSH fallbacks, the fable-terminal build, Pi deployment, dogfood loop, and multi-vendor courier side-tasks.

e.g. Its terminal runbook copies the tools and the worked project.env/worker prompt, queues `fable-terminal`, then starts the pilot with `./pool.sh 1 fable-terminal`.

YAML — A human-readable data-serialization format used for the machine canonSSH — Secure Shell, used for remote machine offloadCLI — Command-Line InterfaceAPI — Application Programming InterfaceJSON — JavaScript Object NotationPyYAML — The Python YAML parsing libraryTUI — Terminal User InterfaceL — Lesson/scar identifierKB — Knowledge-base pattern identifier⚠ — The kit's marker for an assumption, uncertainty, or deferred claim

MDINTAKE.md8.3KB · 2026-07-14 · 11:14 AM

This is the narrative intake questionnaire that turns an underspecified project into a buildable fleet contract. It asks for units and slicing, the source-of-truth spec, gates, dogfood and cold-user completion, guardrails, executors, limits, storage, backup, concurrency, reporting, and unattended recovery, while marking blocking questions.

e.g. The questionnaire asks for a countable atom and its ground truth, such as ticket-to-merge commits, chapters-to-accepted files, or SKUs-to-scanner records.

Q- — Prefix for an intake question identifierG-ID — A named build-failing guardrail identifierAPI — Application Programming InterfaceSSH — Secure ShellL28 — The lesson requiring a cold-user product walk, not just green testsP12 — The ask-once principleKB — Knowledge-base pattern identifier⚠ — The uncertainty marker used for assumptions and deferred answers

MDLESSONS.md49.5KB · 2026-07-23 · 10:16 AM

This is the human narrative companion to the append-only scar log in kernel/lessons.yaml. It records real incidents, the rule learned from each incident, and the structural fix or related practice, so later projects inherit hard-won safeguards instead of repeating fleet-scale mistakes.

e.g. The scar named L29 records a priority unit receiving zero work because its executor lane was already capped, and turns that incident into the rule to probe each lane before launching it.

L — Lesson/scar identifierPROV — Provisional local lesson identifier before canonical numberingEL — Entabeni-kit lesson identifierKB — Knowledge-base pattern identifierYAML — The data format used by the canonical scar fileL34 — The canonical-owns-the-numbers scar about single-writer lesson allocationL96 — The one-standing-harvester scar about preventing concurrent canonical writes

MDNOTEBOOKLM_PROMPTS.md8.5KB · 2026-07-29 · 8:21 AM

This is the prompt pack for regenerating the kit's visual presentation suite in NotebookLM after the textual canon freezes. It defines an exact 12-slide deck, a minimal infographic, and a nine-branch mind map, then gives pinned terminology and failure conditions for QA of generated text.

e.g. The slide prompt requires the tool page to contain exactly the seventeen named tools and the ladder `ollama-local → grok-cli → codex-cli → in-family implementer → conductor → ceiling`.

QA — Quality assuranceP1-P12 — The twelve principle identifiersL1-L29 — Lesson/scar identifiers in the prompt's fixed vocabularyCT — Context-thrift rule familyBC — Behavioral contractKB — Knowledge-base patternTOC — Table of contentsJSON — JavaScript Object Notation

MDNOTEBOOKLM_SOURCE.md25.6KB · 2026-07-29 · 8:21 AM

This is the dense canon source loaded into NotebookLM for independent grounding of the kit's doctrine and feature map. It consolidates principles, lifecycle phases, orchestration layers, vendor seats, tools, templates, integrity rules, terminal/Pi profiles, and the honesty ledger in source-structured prose.

e.g. Its lifecycle section names Phase 2 as the pilot and requires the gate FAIL→fix→PASS cycle, real atoms, control-bus liveness, an honest milestone, and a pause/resume drill.

L — Lesson/scar identifierP — Principle identifierCT — Context-thrift identifierBC — Behavioral contract identifierKB — Knowledge-base identifierCLI — Command-Line InterfaceSSH — Secure ShellTUI — Terminal User InterfaceJSON — JavaScript Object NotationYAML — The machine-readable data format used by the kit

MDORCHESTRATION.md11.8KB · 2026-08-08 · 12:18 AM

This is the human narrative lifecycle for running a project as a fleet: intake, apparatus, pilot, fan-out, watch, done-audit, completion, and optional ship. It also documents recovery keys, machine and quota fallbacks, the self-correction loop, side-task delegation, and the live-testbed feedback loop for deployed fable units.

e.g. The pilot must inject a benign directive that writes a marker, receives an ACK, and continues, because the control bus is the only steering wheel available at scale.

P1-P12 — The ordered principlesL — Lesson/scar identifierKB — Knowledge-base pattern identifierSSH — Secure ShellAPI — Application Programming InterfaceBC — Behavioral contract casePROV — Provisional local scar prefixACK — Acknowledgement from a worker that it executed a directiveRESUME — The on-disk handoff/state record

MDPRINCIPLES.md4.1KB · 2026-07-03 · 1:22 PM

This is the human-readable statement of the kit's twelve ordered principles. It explains why ground truth beats claims, why judgment belongs to the orchestrator, why gates and pilots matter, how to fix systemic classes, how to keep honest ledgers, how to recover from death, and when to ask the human.

e.g. Principle 1 says a DONE claim must be checked against commits, files, records, or a running process before it is accepted.

P1-P12 — The twelve ordered principle identifiersDONE — A completion claim or marker, not proof of completion⚠ — The marker for assumed, deferred, or blocked work

MDREADME.md8.1KB · 2026-07-29 · 8:23 AM

This is the kit's orientation document for a newcomer. It explains that kernel/*.yaml is canonical, the Markdown files are human companions, the fleet method came from a live 29-app build, and the repository contains intake, lifecycle, skills, templates, tools, a terminal blueprint, and a Pi deployment profile.

e.g. Its boot sequence starts by loading `kernel/principles.yaml` and `kernel/lessons.yaml`, then walking intake before instantiating project.env and the pipeline templates.

YAML — A human-readable data-serialization formatMD — Markdown document formatCLI — Command-Line InterfaceLLM — Large Language ModelPyYAML — Python's YAML parsing libraryAPI — Application Programming InterfaceSSH — Secure ShellP — Principle identifierL — Lesson/scar identifierKB — Knowledge-base pattern identifier⚠ — The uncertainty marker

MDUSAGE.md21.0KB · 2026-08-08 · 12:18 AM

This is the longer reference manual for operating a fleet from a populated build directory. It gives quick starts, environment variables, command examples, control-bus procedures, cap and machine fallbacks, terminal/Pi workflows, dogfood expectations, and the full multi-vendor side-task courier protocol.

e.g. For a side-task it writes `.packet.md`, runs `./dispatch.sh start grok AUD-003 /path/to/workdir`, polls bounded `wait` calls, then normalizes the report and re-runs one cited command.

YAML — The machine-readable format parsed for the control busJSON — JavaScript Object NotationSSH — Secure ShellCLI — Command-Line InterfaceAPI — Application Programming InterfacePTY — Pseudo-terminalTUI — Terminal User InterfaceL — Lesson/scar identifierKB — Knowledge-base pattern identifierP12 — The ask-once intake principle⚠ — The uncertainty/deferred marker

MDfeature-map.md13.2KB · 2026-07-29 · 8:21 AM

This is the Markdown companion feature map: a compact, presentation-structured overview of doctrine, lifecycle 0–6, five-layer orchestra, model seats, seventeen headline tools, templates and control bus, integrity, terminal/Pi, and honesty. It is intended to be loaded with NOTEBOOKLM_SOURCE.md as the same-facts visual canon.

e.g. Its model-seat table distinguishes NotebookLM as a verdict oracle rather than an executor and lists `validated / unverified / contradicted` as its outputs.

P1-P12 — Principle identifiersL1-L119 — Lesson/scar identifiersCT1-CT10 — Context-thrift identifiersBC-* — Behavioral-contract evaluation casesKB-* — Knowledge-base patternsCLI — Command-Line InterfaceSSH — Secure ShellJSON — JavaScript Object NotationYAML — The kit's machine-readable data format⚠ — Uncertainty marker

Examples

2 files — worked instances from live runs

ENVexamples/fable-terminal-build/project.env2.2KB · 2026-07-12 · 1:59 PM

This is a worked project.env instance for building the fable-terminal blueprint with the kit. It supplies absolute roots, queue and done-file locations, Codex worker settings, atom-count commands, gate and completion contracts, and the paths needed for the example's single-unit pilot.

e.g. It sets `PROJECT_NAME=FABLE-TERMINAL`, queues the single `fable-terminal` unit, and uses `bun run gate` as the unit's gate command.

ENV — Environment-variable configurationCLI — Command-Line InterfaceYAML — The format used by the instantiated pipeline filesGit — Distributed version-control systemJSON — JavaScript Object NotationL — Lesson/scar identifier

MDexamples/fable-terminal-build/worker-prompt.md3.5KB · 2026-07-12 · 2:04 PM

This is the worked agent-facing prompt for the fable-terminal build worker. It grounds the worker in the canonical spec, guardrails, a single unit workspace, one-shot execution, control-bus checkpoints, atom evidence, completion/product-walk requirements, and the local-model orientation needed for the terminal's own recursive build.

e.g. It tells the worker to read `build-fable-terminal/SPEC.yaml`, implement FT-T atoms in order, run `bun run gate`, and record a PRODUCT WALK before claiming completion.

YAML — The structured agent-facing formatFT-T — Fable-terminal ticket identifierBC — Behavioral contractKB — Knowledge-base patternACK — Directive acknowledgementGit — Distributed version-control systemL — Lesson/scar identifierG- — Build-failing guardrail identifier

Sub-kit · build

6 files — the fable-terminal build program

MDbuild-fable-terminal/ARCHITECTURE.md5.2KB · 2026-07-12 · 1:57 PM

This is the human architecture narrative for fable-terminal, keyed to SPEC.yaml, TOOLS.yaml, BACKLOG.yaml, and EVAL.yaml. It describes the model/tool/permission/memory/compaction loop, maps the implementation to ai-spine layers L0–L7, defends Bun plus TypeScript and strict tool grammar, and states what v1 deliberately omits.

e.g. The central loop is `user turn → system prompt + memory + transcript → model → permission gate → tool result → final text → budget check`, with compaction when needed.

L0-L7 — The eight AI-spine layers from base model through governanceLLM — Large Language ModelTUI — Terminal User InterfaceREPL — Read-Eval-Print LoopAPI — Application Programming InterfaceHTTP — HyperText Transfer ProtocolBC — Behavioral contractEVAL — Evaluation suitev1 — Version 1

YAMLbuild-fable-terminal/BACKLOG.yaml20.9KB · 2026-07-12 · 1:59 PM

This is the executable ticket queue for building fable-terminal under the kit's own pipeline. It groups FT-T tickets into epics covering the core loop, tools, constitution/memory/context, headless orchestration, REPL polish, evaluation, gated hardware seats, persona, and corpus portability, with acceptance conditions for each ticket.

e.g. FT-T009 requires a router that visibly hops on a 429, parks a capped tier with a retry time, uses local fallback offline, and never silently swaps models.

YAML — The agent-facing data formatFT-T — Fable-terminal ticket identifierREPL — Read-Eval-Print LoopAPI — Application Programming InterfaceNPU — Neural Processing UnitEVAL — The evaluation/ship-gate suiteBC — Behavioral contractL7 — The governance layer of the AI spine429 — HTTP status commonly used for rate limitingPTY — Pseudo-terminal

YAMLbuild-fable-terminal/EVAL.yaml10.8KB · 2026-07-12 · 1:57 PM

This is fable-terminal's canonical ship gate, split into model-independent conformance and reference-compared efficacy. It runs headless scenarios in disposable Git fixtures, checks behavioral-contract cases such as empirical verification, recovery, delegation, cascade honesty, offline Pi diagnosis, and product walks, then reports task-class scores against a Claude-Code shim.

e.g. EV-PRODUCT-NOT-PROJECT makes a unit with green tests but a broken fresh-profile flow fail because the agent must perform a cold user walk and name the broken USER_FLOW.

EVAL — Evaluation suiteBC — Behavioral contractEV- — Conformance scenario identifierEF- — Efficacy task-class identifierAPI — Application Programming Interface429 — HTTP rate-limit statusJSON — JavaScript Object NotationPTY — Pseudo-terminalL28 — The green-is-not-shippable lesson requiring a cold-user product walkHTTP — HyperText Transfer ProtocolGit — Distributed version-control system

MDbuild-fable-terminal/README.md4.1KB · 2026-07-12 · 1:57 PM

This is the newcomer-facing README for the fable-terminal blueprint. It makes the key honesty split explicit: the operating layer is reproducible harness behavior, while the model's capability ceiling must be measured side by side against a reference, and it gives the read order and quickstart commands.

e.g. After the backlog is built, the intended use is `./fable` for a REPL, `./fable -p "fix the failing test"` for a pool-launchable worker, and `./fable eval --model qwen3-coder-next` for the report.

README — Repository orientation documentREPL — Read-Eval-Print LoopCLI — Command-Line InterfaceLLM — Large Language ModelYAML — The agent-facing data formatBC — Behavioral contractEVAL — Evaluation suiteSSH — Secure ShellPi — Raspberry Pi single-board computer

YAMLbuild-fable-terminal/SPEC.yaml14.8KB · 2026-07-12 · 1:57 PM

This is the canonical machine-readable product and architecture specification for fable-terminal. It defines the Bun/TypeScript stack, tiered backend router, offline knowledge bundle, memory and context governor, tools and permission modes, headless/subagent orchestration, governance floor, parity targets, and explicit limits on efficacy claims.

e.g. The router falls back from cloud tiers to local Ollama when offline, shows an OFFLINE badge, and requires `kb_lookup` before answering operations or domain questions from the bundled canon.

YAML — The machine-readable data formatLLM — Large Language ModelAPI — Application Programming InterfaceHTTP — HyperText Transfer ProtocolNPU — Neural Processing UnitRAM — Random-Access MemoryREPL — Read-Eval-Print LoopCLI — Command-Line InterfaceBC — Behavioral contractL0-L7 — AI-spine layer identifiersGA-SPINE-LOCK — Guardrail-admin rule preventing the agent from editing its own governanceQ4 — Four-bit model quantizationTTFT — Time to first tokenOOM — Out of memoryTUI — Terminal User InterfaceG-DOCS — The documentation-verification guard identifier

YAMLbuild-fable-terminal/TOOLS.yaml4.6KB · 2026-07-12 · 1:57 PM

This is the contract catalog for fable-terminal's model-visible tools. For each tool it defines parameters, safety class, behavior, path and error conventions, read/write/edit/bus/memory/agent operations, and the permission-gate modes that determine when actions prompt or are automatically allowed.

e.g. The `edit` tool must fail loudly on zero or multiple matches unless `replace_all` is requested and must return a unified diff.

YAML — The machine-readable tool-contract formatAPI — Application Programming InterfaceCT9 — The context-thrift rule that pushes bulk reading to cheaper subagentsBC — Behavioral contractREPL — Read-Eval-Print LoopOFFLINE — The local-only operating mode with no cloud routeJSON — JavaScript Object Notation

Sub-kit · Pi

5 files — the Raspberry-Pi finish program

MDpi-fable-terminal/FINISH_RUNBOOK.md6.8KB · 2026-07-12 · 2:00 PM

This is the self-contained recovery and completion runbook for the reference Pi deployment, written while the Pi was off-network. It begins with an audit of the repo, bus, commits, gate, and corpus state, then gives a clean-worktree ship procedure, bounded corpus ingest, post-deploy probes, live-testbed encoding, and the final owner ping.

e.g. The runbook requires `MemoryMax=1G` for resumed ingest and says cataloging must reach all 71,051 books even if full-text extraction stops at the 20G budget.

Pi — Raspberry Pi single-board computerSSH — Secure ShellGit — Distributed version-control systemACK — Directive acknowledgementOOM — Out of memoryRSS — Resident set sizeG — Gigabyte unit in a corpus budgetBC — Behavioral contractEVAL — Evaluation suiteKB — Knowledge-base retrieval toolL — Lesson/scar identifierd61-d63 — Deployment directive/round identifiers

YAMLpi-fable-terminal/PI_SPEC.yaml11.2KB · 2026-07-14 · 1:48 AM

This is the canonical deployment profile for running the same fable-terminal binary on an 8GB Raspberry Pi 5, with an expanded 16GB/Hailo profile described in the same file. It specifies hardware and OS envelopes, model presets, RAM/thermal management, watcher/hybrid/edge topologies, systemd service settings, hardware upgrades, honest evaluation expectations, and first-session conformance.

e.g. The hybrid_offload topology keeps tools, memory, and ledgers on the Pi while borrowing a LAN model through `OLLAMA_HOST`, visibly degrading to a local model when the LAN box sleeps.

YAML — The machine-readable deployment profilePi — Raspberry Pi single-board computerCPU — Central Processing UnitGPU — Graphics Processing UnitNPU — Neural Processing UnitRAM — Random-Access MemoryLPDDR4X — Low-Power Double Data Rate 4X memoryOS — Operating SystemLTS — Long-Term SupportQ4 — Four-bit quantizationtok/s — Tokens per secondOLLAMA_HOST — Ollama server endpoint environment variableTTFT — Time to first tokenOOM — Out of memoryPCIe — Peripheral Component Interconnect ExpressNVMe — Non-Volatile Memory Express storageSD — Secure Digital storage cardUSB — Universal Serial BusPSU — Power Supply UnitTOPS — Tera Operations Per SecondINT4 — Four-bit integer inference formatL28 — Cold-user first-session conformance lessonL21 — The duplicated-rules-drift lessonEVAL — Evaluation suite

YAMLpi-fable-terminal/RASPBIAN_KNOWLEDGE.yaml7.4KB · 2026-07-17 · 7:58 PM

This is the Pi-specific domain pack rendered into fable's constitution when `profile=pi`. It records Raspberry Pi OS identity, first-response diagnostics, power/thermal/throttle bitfields, boot and firmware paths, PCIe/device-tree, cgroups, GPIO, zram, systemd, Pironman fan behavior, common failures, and non-negotiable safety rules.

e.g. It decodes `0x50005` as undervoltage and throttling happening now plus past undervoltage/throttle history, so benchmarking must stop and the PSU must be fixed first.

YAML — The machine-readable Pi knowledge formatPi — Raspberry Pi single-board computerOS — Operating SystemLTS — Long-Term SupportCPU — Central Processing UnitPCIe — Peripheral Component Interconnect ExpressGPIO — General-Purpose Input/OutputDSI — Display Serial InterfaceNVMe — Non-Volatile Memory ExpressSD — Secure Digital storage cardPSU — Power Supply UnitUSB — Universal Serial BusOOM — Out of memoryRAM — Random-Access MemoryDRM — Direct Rendering ManagerOLED — Organic Light-Emitting DiodeTrixie — Debian 13-based Raspberry Pi OS releasezram — Compressed RAM-backed swap

MDpi-fable-terminal/README.md2.8KB · 2026-07-12 · 1:59 PM

This is the human introduction to the Raspberry Pi 5 deployment profile. It explains that the profile is not a fork, describes the honest watcher, hybrid, and edge-worker seats, points to the Pi constitution pack and setup runbook, and states why an 8GB CPU-only board is useful but not a coding workstation.

e.g. The watcher seat runs a 1.7–4B local model headlessly under systemd to read STATUS, check Git ground truth, decode thermal/power state, and notify the expensive orchestrator.

Pi — Raspberry Pi single-board computerCPU — Central Processing UnitLAN — Local Area NetworkSSH — Secure ShellOS — Operating SystemYAML — The agent-facing file formatL21 — The lesson against duplicating rules in a derived profile

YAMLpi-fable-terminal/SETUP.yaml5.4KB · 2026-07-12 · 1:59 PM

This is the ordered, agent-executable Pi provisioning runbook. Steps S1–S12 install and verify the OS, cooling/power, storage, zram, dependencies, Ollama, models, fable, the Pi profile, the watcher service, and the nightly eval gate; upgrade steps cover the 16GB/Hailo rig and a LAN hybrid.

e.g. S2 requires a 27W PSU and a five-minute CPU stress test whose `vcgencmd get_throttled` result is exactly `0x0`.

YAML — The machine-readable setup formatPi — Raspberry Pi single-board computerOS — Operating SystemPSU — Power Supply UnitNVMe — Non-Volatile Memory Express storagePCIe — Peripheral Component Interconnect ExpressSSH — Secure ShellLAN — Local Area NetworkNPU — Neural Processing UnitHailoRT — Hailo runtime libraryEVAL — Evaluation suiteRAM — Random-Access Memoryzram — Compressed RAM-backed swapUSB — Universal Serial Bus

The God Kit

Commandsclick a command to watch its firing sequence

/add-tour

/animate-it

/backlog

/begin

/book-it

/boot-cost

/check-docs

/close-it

/continue

/create-company

/debrief

/deploy

/docker-purge

/dogfood

/dogfood-it

/drift

/eta

/explain

/fuck-up

/ground

/guardrails

/handoff

/harvest

/hire

/how-much

/learn

/machines

/overnight

/pi

/refactor

/remember

/run-it

/scratch-ass

/ship

/simp-it

/skin-it

/slopless

/small-window

/test-it

/verdict

/watch-local

Every filewhat it is · an example · acronyms defined

Kernel

Skills

Tools

Templates

Root docs

Examples

Sub-kit · build

Sub-kit · Pi