Claude Skills by dwarvesf

Delta log only; the spec carries the design.

Generated: 2026-07-31 Status: APPROVED (operator pre-approved design; adversarial validate overridden in the gate ledger) Lane: full Phase B of the feature-registry program (Phase A: SPEC-219, merged as PR #323). The registry generator makes feature inventory drift mechanical to detect, but nothing yet closes the loop against the hand-derived `docs/workflow-paths.md` path index: a feature can be live in `docs/FEATURES.md` yet missing from the path map, or linger in the path map after removal....

Behavioral proof per `proof-gate.sh contract`'s bar for this task type (`class=behavioral`, no deploy/migration/schema keyword in the diff): the real primary flow (`tests/test-meta.sh`) run end to end, with a negative control on the one test-file assertion this branch touches.

| AC | Claim | Status | |---|---|---| | AC-1 | SKILL.md with frontmatter + four-slot table | MET | | AC-2 | Tier 1 mechanical: freshness refusal + both-directions cross-check | MET (live run below) | | AC-3 | Tier 2 delta-only subagent, no delta = no dispatch | MET (stated; live run had zero delta, zero dispatch) | | AC-4 | PR-gated apply | MET (this PR is the gate) | | AC-5 | README table + header count, audit-loop Known instances | MET | | AC-6 | BACKLOG: ID-458 shipped row + ID-407 registry-s

> GENERATED by `lib/gate/proof-table-gen.sh` from the gate/run ledger (`~/.local/state/dwarves-kit/logs/runs/feature-map-skill.log`). Do NOT hand-edit -- regenerate instead. Companion run-table, not the canonical `proof-of-done.md` (SPEC-016: generators write under `docs/verification/generated/`, never the canonical). Lane: full

CLASS-LEVEL skills: each a rich SKILL.md (plus optional `references/` for session-specific detail), matched by a future agent on its `description`. One broad umbrella with labeled subsections beats five narrow one-session siblings. A library of hundreds of narrow one-bug skills is a FAILURE. - The user corrected your style / tone / format / verbosity / approach. Frustration ("stop doing X", "too verbose", "just give me the answer", "you always do Y") is a FIRST-CLASS signal. - The user correc...

**Merge policy:** auto **Time budget:** 3-4 hours. **Proof:** run-table: the skill fires on "show me the ledger state / my debt / telemetry / token cost" (trigger match) · it queries via the 02 `ledger` CLI (not a re-implementation) · it renders IN-TERMINAL using bot-reply-formatting (tables/bars) · it renders a web ARTIFACT (claude.ai) for share/review · negative control: BOTH surfaces render from the SAME 02 query output (no divergent second data path). Capture a terminal-render sample + an...

Generated: 2026-07-04 Status: VALIDATED Lane: normal (lane-classify returned tiny; escalated per goal-file explicit /spec+/spec-validate mandate, see impl-notes) Depends-on: SPEC-127 (the `ledger` CLI, shipped, PR #673, merged to main as e6ff875b)

> Per-feature record. The canonical multi-feature index is > [`../../proof-of-done.md`](../../proof-of-done.md); this file is its `render-skill` > feature detail. | | | |---|---| | **Profile** | data/CLI tool addition (behavioral, read-only, pure-function render layer) | | **Proof class** | data-tool (recorded live run + negative control + reproducible) | | **Spec** | [`../../specs/SPEC-128-render-skill.md`](../../specs/SPEC-128-render-skill.md) |

Use when auditing, reconciling, or verifying a repo's `_meta/BACKLOG.md` Active queue against reality, "audit the backlog", "reconcile the board", "is the board still true", "dọn backlog", "rà soát backlog", "check board rows against their specs", or when a scheduled run asks for the backlog-reconcile audit of a repo. NOT for filing new items onto the board (board-registration tooling, e.g. an adopter's own intake skill), NOT for a single known-stale row (just fix it), NOT for this kit's own ...

Use for the whole-estate CI audit, "audit the CI", "are the runners clean", "check the release pipeline", "ci drift", "stale workflows / secrets / runners", or a scheduled CI-audit cadence run. Enumerates every workflow + GitHub-side CI state (enabled/disabled, secrets, vars, runners, releases, environment policy), verdicts each against the live repo with evidence, fixes drift on a branch, gates through a PR. An audit-loop instance (docs/patterns/audit-loop.md). NOT for one failing run (that ...

Use for the whole-estate doc audit, "run the doc-drift loop", "audit the docs against the code", "are the docs still true", "doc drift sweep", "check every doc claim", or a scheduled doc-audit cadence run. Enumerates every LIVING doc (README, MANUAL, AGENTS, WORKFLOW, architecture, patterns), verdicts each against the live repo with evidence, fixes drift on a branch, gates through a PR. An audit-loop instance (docs/patterns/audit-loop.md). NOT for diff-scoped doc sync inside a build cycle (th...

Use to audit the gauntlet records, "are the ROUNDS.md records honest", "verify the proof corpus", "check each run record against its evidence", "audit the gauntlet corpus", "is the gauntlet proof trustworthy", or a scheduled gauntlet-proof-audit cadence run. Enumerates every committed gauntlet run record (ROUNDS.md/AB-ROUNDS.md), verdicts each claim against its own committed evidence (markers, checker-output.txt, scrub, run-dir grammar), reports discrepancies. An audit-loop instance (docs/pat...

Fetch curated API documentation using Context Hub (chub) before coding against any external API. Use when the task involves calling a third-party API (Stripe, OpenAI, Airwallex, Notion, etc.) to prevent hallucinated parameters and deprecated endpoints.

Use when the user wants to design or add a new bounded loop to the kit's own SDLC orchestration ("let's build a loop", "design a new loop for the orchestrator", "loop engineering", "make this a bounded loop", "is this worth a loop"), or mentions the Karpathy loop / autoresearch / hill-climb / search-and-select prompt or parameter tuning. Walks the gate (should this even be a loop) then the anatomy (artifact / scanner / reviser / stop condition), routing to one of three shapes: bounded-revise ...

Use when auditing, consolidating, or cleaning a repo's .claude/memory store, "dọn memory", "memory tidy", the biweekly memory audit, duplicate or stale memory notes, MEMORY.md index drift (notes missing from the index, index entries with no file), or when a scheduled run asks for the memory-tidy audit of a repository. NOT for the built-in machine-local auto-memory under ~/.claude/projects (not git-tracked, no PR gate) and NOT for editing a single note (just edit it).

Query and render the kit's control plane from an agent session. Use when asked about the fleet's runs, gate verdicts, conformance, spend/tokens/cache economics, the cognitive-debt score, replaying a recorded run, benchmark results, or building/refreshing the observability dashboard. Trigger phrases include "dashboard", "control plane", "how much did we spend", "token usage", "cache hit rate", "cognitive debt", "debt score", "replay run <rid>", "gate log for <rid>", "conformance", "fleet stats...

Use for the whole-repo decay audit, "run the repo-hygiene loop", "what has rotted in this repo", "audit repo hygiene", "find the stale files", "is anything mis-shelved", "dọn repo", or a scheduled repo-hygiene cadence run. Enumerates files inside a git repo that have decayed (unreferenced docs, stale staging drops, records parked in a control surface, logs past their documented budget, large cold gitignored dirs), verdicts each with evidence inline, co-locates mis-shelved records on a branch,...

Review and promote skill drafts that skill-curator staged from past sessions. Use when the user runs /skill-review, says "review my skill drafts", "promote a staged skill", or asks what skills the self-improvement loop proposed. Lists drafts under ~/.claude/skill-proposals/, runs the writing-skills quality bar on each, then promotes the approved ones into ~/.claude/skills/ or rejects them.

Query or render the state of the scattered kit/tide/tg-cleanup/learned ledgers (the dwarves-kit gate/proof/telemetry corpus, tide file-move state, tg-cleanup snapshots, the learning ledger), and detect + propose backlog rows off anomalies in that state. Use when the operator asks to SEE ledger state -- "show me the ledger state", "my debt", "understanding debt", "telemetry", "token cost" / "how much am I spending on tokens", "kit runs" / "kit lane telemetry", "ledger status", "render the ledg...

Maintainer-only (dwarves-kit repo dev only). Use to audit THIS KIT's OWN feature estate against its path map, "run the topology-drift loop", "is the workflow path map still complete", "does every feature appear in workflow-paths", "feature inventory drift", or after a batch of merges that added/removed commands, agents, skills, or hooks. Cross-checks the generated docs/FEATURES.md registry against the docs/workflow-paths.md path index both directions, then re-places only the DELTA features on...

Use for the live-website agent-readiness audit, "run the web-drift loop", "audit our websites", "are our sites still readable by an agent", "web drift sweep", "check the public sites", or a scheduled site-audit cadence run. Enumerates every site the consumer declared in WEB_DRIFT_SITES, probes each over read-only HTTP with lib/webcheck (groundwork, page, and API tiers), verdicts each check with evidence, and files the fixes as board rows in the repo that owns the site's source plus a report. ...