A curated collection of practical, evidence-backed skills for coding agents.
This repository also configures its own agents. See what agent-skills tells them →
Skill Codex
Audit, score, compare, report, and iteratively improve how safely and effectively coding agents can work in a software repository. Use for read-only agent-readiness audits, first-class HTML/PDF readiness reports, readiness levels or percentages, Factory-compatible comparisons, AGENTREADINESSPREFERENCES.md setup…
Skill Codex
Enforce a TypeScript policy that bans as type assertions and replace casts with compiler-verified narrowing or runtime validation. Use when introducing the ESLint rule, removing violations, reviewing assertion workarounds, or designing typed data boundaries.
Skill Codex
Perform an evidence-backed adversarial review of a diff, commit, branch, pull request, or named files. Use only when the user explicitly requests a fresh code review or findings on implementation quality; do not use merely to fix CI or act on existing review comments.
Skill Codex
Design, implement, or review a human-first command-line interface and developer workflow. Use when commands are long, flag-heavy, difficult to discover, require manual setup glue, expose implementation paths, need interactive onboarding, mix human and automation concerns, or create avoidable developer-experience…
Skill Codex
Add meaningful tests for changed or recently committed code, run focused and full suites, and fix real implementation defects exposed by those tests. Use when asked to add tests, improve coverage, test new code, or make sure a change is well tested.
Skill Codex
Resolve Knip unused-export findings by classifying each symbol before editing and preferring dead-code deletion, local visibility, or deliberate package contracts over suppression. Use for unused exports, exported types, barrel re-exports, test-only exports, and apparent false positives.
Skill Codex
Classify proposed tests as valuable behavior tests, harmful implementation tests, or temporary proof/TDD tests. Use before writing, expanding, reviewing, or deleting tests, especially during coverage work or when a test would keep questionable code alive.
Skill Codex
Triage findings from automated or AI code-review systems, verify them against current code, and either present a human-gated action plan or auto-fix only narrowly safe correctness issues. Use for review bots, synthetic review PRs, or unattended self-heal workflows.
Skill Codex
Create polished local review artifacts as a Markdown source, self-contained HTML report, PDF, and rendered page images. Use for planning sheets, research summaries, run reports, probe results, or any disposable local document that benefits from visual review.
Skill Codex
Design, run, and debug trustworthy local browser end-to-end tests with isolated ports, disposable data, readiness checks, deterministic seeds, and state restoration. Use for local web smoke suites, authenticated flows, multi-service launchers, or E2E environment hardening.
Skill Codex
Create, review, render, and maintain Mermaid diagrams for architecture, flows, sequences, state machines, schemas, timelines, and tradeoffs. Use when a visual model materially clarifies structure or behavior and the result should remain diagrams-as-code or exported SVG/PNG.
Skill Codex
Explicitly invoked MVP-launch mindset that combines the no-bullshit skill with ruthless focus on the smallest genuinely viable product. Use only when the user unmistakably invokes $no-bullshit-launch, asks to use the no-bullshit-launch skill, or explicitly turns on no-bullshit launch mode. Never trigger automatically…
Skill Codex
Explicitly invoked mode for direct, brutally honest, plain-English, practical collaboration and implementation. Use only when the user unmistakably invokes $no-bullshit, asks to use the no-bullshit skill, or explicitly turns on no-bullshit mode. Never trigger merely because the user asks for brevity, honesty…
Skill Codex
Use Perplexity MCP for current, cited web research. Trigger when the user asks to use Perplexity, asks for deep research, market/vendor/provider research, current comparisons, current API/service capability checks, or wants citation-backed investigation beyond ordinary browsing.
Skill Codex
Review and diagnose Postgres-backed job queues for MVCC dead tuples, index bloat, autovacuum lag, long transactions, claim-query performance, retention, and SKIP LOCKED safety. Use when implementing, debugging, reviewing, or scaling database queues and their dashboards.
Skill Codex
Initialize, plan, run, inspect, and review resumable Ralph-style autonomous implementation loops with sequential chunks, persistent scratchpad memory, manifests, and idempotent post-sprint hooks. Use for .ralph orchestration or similar long-running coding-agent sprint systems.
Skill Codex
Critically review a current or completed Ralph sprint for specification alignment, implementation correctness, validation evidence, orchestration reliability, and remaining gaps. Use when the user asks to review, audit, accept, or assess a Ralph sprint. Do not implement fixes unless the user separately authorizes them.
Skill Codex
Create, repair, validate, or advance a Ralph sprint from a repository specification. Use when the user asks for the first or next sprint, sprint planning, spec breakdown, chunk design, CURRENTSPRINT selection, or preparation after a completed sprint. Never start the autonomous loop unless separately requested.
Skill Codex
Report truthful Ralph loop progress from runtime state, chunks, manifests, hooks, markers, commits, and logs. Use when the user asks what Ralph is doing, whether a sprint or chunk finished, why it stopped, whether it is safe to resume, or what remains. This is read-only unless the user separately requests repair.
Skill Codex
Use when the user wants to clean up, declutter, reorganize, rename, or simplify a repository. Scans the whole repo and produces Potential/STRONG candidates for deletions, moves, and renames, then waits for explicit approval before changing files.
Skill Codex
Inspect a computer and software repository, detect installed coding agents and tool prerequisites, recommend relevant skills or outcome-based packs with evidence, and safely vendor selected skills through the upstream Skills CLI. Use for first-time onboarding, repository-local skill setup, prerequisite diagnosis…
Skill Codex
Use ONLY when the user explicitly asks to roll out, stage, ramp, dry-run/live-run, or operate a script that writes to live data, shared Postgres mirrors, or external systems. Do NOT use for ordinary code edits, reviews, tests, pipeline investigation, bug fixes, script design, or merely noticing that a script could…
Skill Codex
Evaluate an old or explicitly stale pull request against current default-branch behavior, product plans, code quality, and architecture. Use for backlog cleanup and close, salvage, refresh, or merge recommendations; do not close or modify a PR without authorization.
Skill Codex
Prevent unused UI source, CSS growth, Tailwind scanning waste, and frontend dependency bloat while adding or changing components and visual effects. Use for UI primitives, Tailwind configuration, CSS files, component libraries, or frontend dependency changes.
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: