App_Registry.md gap closed, 2026-07-28 — cinderdoc.thewolf.tech, releases.cinderdoc.thewolf.tech, and agentify.thewolf.tech were live per Holt's domain-watch but appeared in no canonical doc. Added a new section documenting all three, sourced from each project's own repo (not inferred from name), and noted explicitly that neither is part of the Holt -ify roster — separate standalone products sharing the same infrastructure.
OPERATION HOLT-ECOSYSTEM-PACKETS
Cards are harvested from this app's BACKLOG/NOTES/TODO/ROADMAP files, handoff tokens, explicitly tagged central backlog lines, and one labelled repo baseline. The live dot, last-activity, and dev/prod drift are computed from disk and the session registry (cached ≤3 min). Central BACKLOG.md sample: Jul 28, 2026 · 3:46:22 PM EDT.
background refresh last succeeded 15s ago · stale after 270s
Backlog
0No cards in this column.
In Progress
20[sensitive content omitted]
/continue holt-grad3 — handoff: close-2026-07-23-conductor.md
/continue ct12 — handoff: ct12.md
/continue observify — handoff: ct12.md
/continue skills — handoff: ct12.md
/continue dusk — handoff: dusk.md
/continue holt-grad2 — handoff: holt-grad2.md
/continue holt-grad3 — handoff: holt-grad3.md
/continue holt-security — handoff: holt-security.md
/continue observer — handoff: observer.md
/continue tokens — handoff: observer.md
/continue dusk — handoff: observer.md
/continue observify — handoff: observer.md
/continue observify-holt — handoff: observify-holt.md
/continue observify — handoff: observify.md
/continue terminus — handoff: terminus.md
[sensitive content omitted]
terminus capstone A/B (holt-ecosystem-packets/_terminus-eval) — HOLT window autonomous tonight · claims 14/17 scoreable security PASS (82%) ⚠ unaudited (file said 10/17) · NEXT: audit the 14/17 claim, fold into AB_SCOREBOARD.md — token: [redacted]
Repo baseline — Holt Ecosystem — Packet Workspace (README.md)
Currently Running
0No cards in this column.
In Development
0No cards in this column.
Testing
0No cards in this column.
Done
2_terminus-eval/harness/score.sh — materialize gap fixed for Holt's practicum, done 2026-07-28. Installed qemu-user-static/binfmt-support on the Pi (previously only binfmt-support was present, with no emulation handler registered — Ruby's amd64-only pinned digest errored exec format error); verified fixed with ruby -e 'puts "ok"' running under emulation and again seeing qemu auto-engage inside a real docker build. Edited the canonical score.sh (backed up first to score.sh.bak-pre-materialize-fix-2026-07-28): extended the existing verifier PATH-restore fix to cover the materialize exec too (it was verifier-only before, so cargo/etc. dropped out of PATH under bash -l during materialize — Go was immune because its Dockerfile symlinks go onto the bare PATH, Rust's doesn't); added a generic Rust materialize hook (finds every Cargo.toml under /app, builds --release --offline with a --locked fallback, installs binaries to /app/bin/<name>, best-effort no-arg run since that's not a universal CLI contract); added a generic multi-stage-Dockerfile leak-cleanup step (parses the Dockerfile itself for context paths a non-final stage COPYs that the final stage never does, strips exactly those from the injected --agent tree — no task hardcoding). Regression-tested rbac-session-activation (Go) and colony-requeening (Python) still PASS; new coverage ecdsa-nonce-reuse-scan (Rust) PASSES including its test_sigforge_source_not_in_agent_image anti-cheat check, and a second Rust task (rust-secret-scope-reduction-planner) also PASSES. All via reward.txt, not exit code. C++/Java hooks were investigated and deliberately NOT added — both real C++ tasks checked have Dockerfiles whose COPY destinations don't align with the harness's fixed /app injection root (subdir flatten/nest mismatches), so a hook would either no-op or silently build the wrong tree; the one Java task in the corpus is milestone-format, which prepare.sh already rejects outright. Flagged rather than forced, per plan. All ephemeral Docker images/containers and the Pi scratch dir were cleaned up afterward; no solution/proprietary task source was left on disk.
think-like-fable kit (holt-ecosystem-packets/think-like-fable = CANON) — 149 scars, 21 skills · shipped tonight (backlog skill + simp-it scoping) to 16 mirrors + 2 zips · NEXT: deliberate /harvest of tonight's 5 candidates
Deployed
0No cards in this column.