Claude Skills by anthony-chaudhary

One repeatable RSI pass over CODE quality — the code-side counterpart of refresh-readme. Runs the code-quality scorecard (tools/code_quality_scorecard.py), reads the code-debt work-list, retires debt worst-first using ONLY the safe, genuine classes (gofmt, real tests for untested packages, safe god-function extraction), re-measures to PROVE the number dropped, grounds the ship in DOS (dos commit-audit on the new commit, dos review for the ship_integrity KPI), and commits by explicit path. Use...

The super-loop-family member that ASKS instead of ships. It launches detached workers whose only job is to ask 5–10 hard, honest questions about what the repo is doing — the question no other agent has asked, the one everyone's afraid to ask, the one that's opposite what the repo claims, the steelman of the other side — and append them to a durable ledger (docs/questions/asked.jsonl). A SEPARATE next-step loop, in a SEPARATE context window, turns qualifying questions into gh tickets. Use when...

Refresh the recent-operational cachevalue numbers in a guarded doc (e.g. docs/integrations/fable5-more-usage-for-free.md) when this-week's telemetry has moved on. Re-derives the frozen snapshots from live `fak cachevalue report`, reconciles the doc's rendered numbers + snapshot_date to the fresh capture, and re-runs the hermetic audit until it is clean. The audit (tools/cachedoc_numbers_audit.py, gated in `make cachedoc-numbers-lint`) binds every rendered number to a committed snapshot field ...

One repeatable pass over README.md — the front door — that keeps ONLY the most important points current and honestly framed. Runs the freshness auditor (tools/readme_freshness_audit.py), turns each FAIL into a required edit and each WARN into a judgment call, applies the three front-page laws (SOTA-vs-us-never-naive, 6th-grade/Feynman-but-accurate, wide-audience), re-stamps the readme-verified marker, and commits ONLY README.md (+ the tool) by explicit path. Use after a release / version bump...

Perform a full versioned release — bump version, draft release notes, commit, tag, push, and create the GitHub release page. Reads `.claude/project.yaml` for the project's release-context and version-bump helpers; the skill text is universal, the helpers are project-supplied. Use when the user says "cut a release", "ship vX.Y.Z", "release", or after a shippable phase.

Audit and recover crashed Claude and Codex sessions through one dry-run-first cohort surface, with exact provider identity and post-launch transcript/thread advancement; then audit the scheduler tower that keeps automatic recovery alive. Use when this named workflow matches the task.

Plan and run one bounded overnight issue worker through fak's guarded dispatch path, with typed capacity admission, lane leases, explicit dry-run/live gates, and independent git/DOS/test reconciliation. Use when the operator asks to let a narrow repo task run unattended overnight.

The generic 2×-then-discover loop the scorecard family runs — enumerate exhaustive raw debt, drive it down 2× with genuine fixes, rescore to PROVE the drop, and continuously harden discovery under a new metric version without capping debt or rewriting history. The conductor over the per-surface instruments (quality-score, industry-score, persona-score). Use when this named workflow matches the task.

The generic scoring doctrine the whole fak scorecard family instantiates — how to BUILD a new deterministic, tree-cross-checked, debt-driving scorecard and how to RUN any of them as a repeatable RSI pass. Every sibling (code-quality, docs, doc-appeal, seo, demo-quality, repo-hygiene, observability, learning, industry, agent-readiness, product, persona) is the same machine pointed at a different surface: pure KPIs over a data-dir or the git-tracked tree, cross-checked against reality so the sc...

The super-loop that closes the research→backlog loop — it chains the outward CRAWLERS (the daily `idea-scout` arXiv/GitHub feed, the industry scans, the RESEARCH/CONCEPT corpus) into the STUDY pipeline (`/study-repo` → `/field-borrow`) and runs the whole thing on a cadence. The crawler surfaces repo-shaped leads into a needs-triage queue and stops; turning any one into scoped, witnessed, license-clean backlog is still a manual pass someone has to remember to run. This skill is that seam, auto...

Shape an answer for fast scanning with result-first sentences, concrete verbs, compact structure, and selectable compression while preserving every correctness-critical detail. Use when the user asks for Caveman-style output, terse or low-token prose, executive scanning, fewer words, or a concise rewrite whose substance must remain intact.

Witnessed lifecycle for the project skill pack — usage-telemetry sidecar, value/idle-driven auto-archive (never delete, restorable), pin-exemption, journaled reversible transitions. Use to record skill usage, review archive verdicts, archive or restore a skill, or pin one exempt. Max action is archive; nothing is ever deleted. Use when this named workflow matches the task.

Flag content-redundant SKILL.md pairs as merge candidates — pairwise body cosine similarity over the skill pack, sorted highest-first. Use when the pack feels bloated or two skills seem to overlap, to find "skills A and B are N% token-similar — consider merging" candidates before a skill-lifecycle archive pass. Proposal only — never merges, edits, or deletes a skill.

One repeatable pass that keeps the skill pack itself effective — the one surface no other scorecard grades. Runs the skill-effectiveness scorecard (`fak skill-effectiveness-scorecard`) over every .claude/skills/*/SKILL.md, reads the skill-debt work-list, and retires it worst-first by ADDING the real affordance — a sharp "Use when …" trigger, a reference that resolves on disk, the commit-by-path discipline a committing skill owes the shared trunk, a witness step, a scoped allowed-tools — never...

One repeatable RSI pass over CODE SLOP — the slop the compiler can't see. Runs the code-slop scorecard (tools/code_slop_scorecard.py), reads the slop-debt work-list, retires debt worst-first using ONLY genuine fixes (extract a copy-paste clone into a shared helper, delete or wire a dead unexported symbol, add a real assertion to a vacuous test, drop tautological doc comments + commented-out code), re-measures to PROVE the number dropped, regenerates the committed snapshot, grounds the ship in...

One repeatable pass that stops fak from re-inventing known kernel art - before writing or optimizing a compute kernel (a quantized GEMM, a fused attention, a KV-cache reuse, a MoE dispatch, a Metal/CUDA kernel), it checks the SOTA prior-art matrix for the production reference (llama.cpp / Marlin / CUTLASS / FlashInfer / vLLM / SGLang / a named paper), decides the route (borrow / bind / stay-minimal), holds the result to the named oracle, and records what was consulted in. Use when this named ...

Apply the two new-work defaults — ship the minimal WORKING end-to-end spine first (or file the spine as its own issue), then fan out the 3..50+ follow-on QA/dogfood/productization backlog at creation time via `fak issue fanout`. Use when starting any new feature/leaf/verb/demo, when a spine just shipped, when asked to "fan out", "file follow-ons", "create the e2e spin", or at the end of a super-loop turn that landed new work.

One repeatable pass that keeps fak trustworthy while it iterates fast — the question no other scorecard asks: as we add items quickly, how do we KNOW a regression / tail-wag / confusion landed, and how do we REVERT to a stable version? Runs the stability scorecard (tools/stability_scorecard.py) over the git-tracked tree across four groups — sentinel (a regression turns a gate RED), invariant (the core assumptions are encoded as tests), revert (we can roll back: keep/revert ladder, version pin...

Turn a fak stale-work packet into dedicated, contract-valid issue units, collision-safe dispatch waves, and witness-only reconciliation. PLAN by default; use when an operator asks to adjudicate stale-work candidates, file the dedicated issues, or launch fresh workers for already-filed stale-work issues.

The operator loop over the steer-prs overlay (`fak steer prs`) — fak's read-only view that folds the pending dev->release trunk delta into PR-sized units per (fak <leaf>) ship-stamp and renders them WORST-ATTENTION-FIRST (RESIDUAL -> UNVERIFIABLE -> CLEARED). Teaches the four-step loop (run the view and read worst-first; apply the REGIME GATE — a CLEARED unit with a healthy curve is a reason to do nothing; pick the WEAKEST SUFFICIENT RUNG on the observe -> comment -> ack -> redirect ->. Use w...

One repeatable pass that keeps fak as STEERABLE as it grows — the one scorecard whose every KPI is growth-invariant, so a 2x-larger repo with the same discipline scores the same. Runs the steerability scorecard (tools/steerability_scorecard.py) over the working tree, reads the 0-100 steerability index + the advisory drift signals (coupling hubs, p90 sizes, long-function rate, package drift, churn hot spots), drives the index UP and the worst drift axis DOWN by adding REAL modularity (split a ...

Inventory and deeply study innovation in upstream pull-request queues, including open and unmerged PRs, then turn selected mechanisms into provenance-honest fak research and deduplicated GitHub issue drafts. Use for vLLM or any high-value repository when useful work may exist before release or merge, when maintainers cannot review a busy queue fast enough, or when asked what fak should learn, borrow, watch, reject, or track from incoming PRs.

High-priority deep study of external code and proposals for fak. Invoke proactively whenever a repository, package, PR, issue, release, paper-with-code, or implementation is relevant—not only on explicit study requests. Acquire into scratch and pin revisions; mine code, tests, docs, history, releases, open and closed issues, PRs, discussions, roadmaps, and license/provenance; date every observation; directly port or adapt implementation when licensing permits; and explore both shipped mechani...

Plan, price, launch, monitor, and reconcile bulk headless issue-resolution work safely. Use when an operator asks for a super loop, worker wave, detached issue workers, backlog draining, capacity/status, stale-worker cleanup, ramp-up, or an overnight fleet. Dispatches end-to-end issue owners, requires explicit launch intent, prices account and tree capacity, verifies effects from git or DOS witnesses, and closes every child, lease, and intent cleanly.

Find the most "tail wagging the dog" part of the system — a peripheral or secondary concern that is disproportionately driving core design decisions. Produces a ranked list of inverted-priority findings with evidence and a proposed rebalance. Use when the user asks to find the "tail wagging the dog", inverted priorities, misplaced drivers, mis-layered concerns, or when something small is clearly dictating something big.

One repeatable pass that decides whether a GitHub ticket is a single dispatchable unit of agent work — or names exactly which of the six scope axes it fails and how to fix it. Wraps the native scope toolkit (`fak issue contract` for structure/size/routing, `fak dispatch issue-smallness-lint` for atomicity, `fak issue cohort` for batch/wave placement) and reads back one verdict per issue: DISPATCHABLE, or TRIAGE (add the missing section), DECOMPOSE (S2+ epic → leaves), or SPLIT (two deliverabl...

One repeatable pass that keeps fak's OUT-OF-THE-BOX token economy amazing — every stacking token-saving method that can SAFELY default is on by default, honestly noted, and locked against regression. Runs the token-saving-defaults scorecard (`fak token-defaults-scorecard`) over the entrypoint source (cmd/fak/guard.go, cmd/fak/serve.go, the gateway Default* constants, and the audited servewiringData rows), turns each HARD defect into a required fix — turn a WITNESSED-safe bounded-loss saver on...

Audit recent Claude and Codex transcript JSONL with the first-class Go `fak trajectory audit` verb: exact token/cache buckets, source coverage, behavior, deterministic bottlenecks, semantic confusion checks, and baseline regressions. Use for cross-session cost, efficiency, repetition, reasoning-quality, and churn questions.

The operator on-ramp to trajectory control (`trajctl`) — fak's live, forward-progress control plane over a DECLARED objective. Teaches the one primitive that carries the family (anything you want to progress gets a named, witnessed score, and every move is either improve-the-score or improve-the-scorer), then the four operator loops: declare an objective + plan + budget, read the CURVE (never a point) for the closed signal vocabulary (HEALTHY / STALL / DRIFT / DETOUR_OVERRUN), apply the when-...

One repeatable gardening pass over a trajectory corpus — the JSONL of per-turn Turn rows a fak trajectory.Recorder exports. Uses the `fak traj` toolkit (the data plane + the simhash reference vector-similarity primitive + the pluggable scorer seam) to find the trajectories worth a human's attention — near-duplicate queries the lexical ranker misses, cost outliers, and traces the kernel kept refusing — then PROPOSES prune candidates (it never deletes). The reference application of fak's trajec...

Bind a done-claim to a GREEN test run of the changed package, not just diff shape. Use after a commit claims a package/feature is done and you want to run that commit's affected tests and report CLAIM_TEST_GREEN / CLAIM_TEST_RED / CLAIM_TEST_UNRUN before folding the claim as true.

The honest closing half of a super loop — after a detached bulk wave (`/super-loop`) has run, HARVEST it: witness what each headless worker actually shipped (not what its log claims), re-queue the leaves that were claimed-done-but-not-shipped, stop workers that are spinning without net gain, and surface any lane a worker stranded dirty. A launch is not a ship, so a bulk loop is only durable if something reconciles its output against git ground truth. Read-mostly — it audits and. Use when this...

Ultra-compressed communication mode. Cuts output tokens 65% (measured) by speaking like caveman while keeping full technical accuracy. Supports intensity levels: lite, full (default), ultra, wenyan-lite, wenyan-full, wenyan-ultra. Use when user says "caveman mode", "talk like caveman", "use caveman", "less tokens", "be brief", or invokes /caveman. Also auto-triggers when token efficiency is requested.

Search repository text through a deterministic first-class fak command.

Compact the context window and prune the memory store so a long session stays under budget. Use when the context is getting large.

Score the repository's code quality and emit a ranked table of findings. Read-only.

Compact the context window and prune the memory store when a session grows too long.

Score the repo quality and emit a report card.

Ultra-compressed communication mode. Cuts output tokens 65% (measured) by speaking like caveman while keeping full technical accuracy. Supports intensity levels: lite, full (default), ultra, wenyan-lite, wenyan-full, wenyan-ultra. Use when user says "caveman mode", "talk like caveman", "use caveman", "less tokens", "be brief", or invokes /caveman. Also auto-triggers when token efficiency is requested.