🔄 Registry-reported update — August 12, 2026. The exact runnable-file count is derived from the registry and recorded inside manifest.json. This wrapper does not independently attest that a daily propagation job ran or that the ZIP matches a published hash; treat the date as directory metadata until an artifact digest and verification receipt are exposed.
Where this sits in the system
A skill is a document. A pack is a folder of documents. Neither one does any work. Work happens when a job runs those skills on a schedule, checks its own output, and keeps its files somewhere it can read them again tomorrow. That is the whole difference between owning skills and having an agent.
- SkillYOU ARE HERE — one task, written down to a standard, so an agent can run it without you in the room. The dated core snapshot contains 239 local task-skill files; the live registry may be larger.
- Pack — those skills bundled into a download you install in one paste.
- Agent — a named role with a job description — not a chat window you retype every morning.
- Job — a schedule, a QA cycle, and somewhere to keep working files. Miss any of the three and nothing runs twice.
- Proof — every finished run written up in public, and the lesson pushed back into the skill.
The map: The System · every asset: Asset Tracker · next door: How we build agents.
Monthly capability refresh — and the first entry this panel has been able to record since it was built
Reviewed the July 2026 releases across Anthropic, Google and OpenAI and rewrote the operating layer against them. Opus 5’s per-turn reasoning-effort dial (July 24) became a routing rung of its own — raise effort before raising model — and every scheduled job now names a fallback tier, after Fable’s two-and-a-half-week suspension in June proved a single-engine routing table is a single point of failure. Sonnet 5 replaces the smaller models as the default for any step that chains tool calls. The same run found three of three download buttons advertising hand-typed counts frozen months earlier: /dealcon offered 10 skills for a 19-skill pack, and the Task Library surfaces mixed a 239-file hub count with larger versioned downloads. This page now labels the live registry, runnable ZIP, and archived bundle separately; upstream PR #1 makes the Task Library counts build-derived after merge. What broke and what we changed.
Gap in the record. One monthly cycle is missing from this record. The panel used to live inside the dashboard app itself, which moved to a host we hold no write credential for — so from June 10 to August 1 there was no way to append to it, while the daily “updated” badge above it kept stamping the correct date. The gap is left visible on purpose. A proof-of-recursion panel that hides the months it could not record is not proof of anything.
Task Library dashboard published
Legacy migration imported 239 local task-skill files across 13 categories. That count did not prove definitive-article coverage, ownership, or production reliability; current status and evidence fields expose those gaps. How the library was built.
Every monthly refresh appends an entry here, and this panel lives in this page’s own content — not inside the dashboard app — so the job that is supposed to write to it always can. The changelog is the proof of recursion; a panel nothing can append to proves nothing.