agent-skills

Use only when explicitly asked to archive/prune/compact/roll over checked tasks from TODO.md into `.ai/todos/TODO_UNTIL_YYYY_MM_DD.md`, leaving unchecked tasks.

Use only when explicitly invoked for Git commit workflows: stage intended changes, craft Conventional Prefix Format messages by default, Natural Language messages with --natural or configured repos, commit, and optionally --all, --staged, --deep, --close, or --push.

Use for dependency updates: update/bump deps, npm/pnpm/yarn/bun package upgrades, outdated checks, package.json updates, or taze.

Use for release versioning: bump/cut/tag a release, bump version, create a release, changelog updates, or version tagging.

Use for GitHub PR/issue/discussion workflows: create/update PRs or issues, post comments, start discussions; triggers include create/open PR, file/update issue, yeet.

Use to polish recently changed code: simplify for readability/maintainability and run a risk-profiled review that autonomously applies fixes. Default runs both passes; pass --simplify or --review for one. Covers code/PR review, audits, bug/security checks, reviewing diffs or changes, cleanup, refactoring, and reducing complexity.

Use for GitHub CLI automation: gh commands, repo info, workflow triggers, GitHub search, codespaces, PR status, issues, repo browsing, or command-line GitHub tasks.

Use for autoresearch, autonomous experiments, optimization loops, "optimize X overnight/in a loop", or "experiment loop"; sets up iterative trials for an optimization target.

Use for just/justfile task automation: create justfiles, write recipes, configure settings, add modules/attributes, or set up command-runner workflows.

Use to open CoinGecko historical-data pages in the default browser for a coin/date or to show historical price on date X.

Use for nontrivial Effect-TS work including services/layers, typed errors, Schema/JSONSchema, Config, runtime/concurrency, @effect/vitest, @effect/ai, @effect/sql, or @prb/effect-next.

Use to retrospect on a task against historical Codex and Claude Code chat transcripts in the current project, identify recurring agent mistakes, and recommend or apply durable fixes such as AGENTS.md updates or new skills.

Use to audit Agent Skills catalogs or installed skill roots: validate SKILL.md frontmatter, Codex openai.yaml metadata, README catalog coverage, cli-* version pins, and relative resource links; optionally apply conservative --fix-safe metadata repairs.

Use for Foundry cast/on-chain actions: send tx, call contract, sign message, cast send/call/wallet, encode/decode ABI/calldata, or RPC interactions. For read-only balances/history, prefer evm-atlas.

Use for targeted EVM chain, account, transaction, RPC, explorer, and bridge data: chain name/ID, native symbol, RouteMesh, balances, token/NFT holdings/transfers, tx history, funding origin via Etherscan/Blockscout/Chainscout; Bungee, LayerZero, LI.FI, Socket enrichment.

Use to find agent skill installs, duplicate skills, cross-dependencies, invocations, and cross-references across the local machine. Scans broadly while ignoring macOS protected home paths, agent home install/state directories (~/.agents, ~/.claude, ~/.codex, ~/.local/state/skills), transcript/history/state directories, dependency folders, and backups.

Use only when explicitly invoked to rename a GitHub upstream repo and matching local project folder, update the local origin remote, and preserve Codex CLI and Claude Code thread continuity references.

Use when CSV, TSV, or Excel (.xlsx) is the primary input/output: inspect, clean, transform, dedupe, merge, validate, convert, recalc formulas, or create/fix spreadsheets. Do not trigger when tabular data is incidental.

Use to create or polish repo agent context files: README.md, AGENTS.md/CLAUDE.md symlinks, and existing project .agents/skills/SKILL.md files. Updates existing skills only; do not create skills.

Use to create/scaffold/init a new agent skill in `.agents/skills` by default or `~/.agents/skills` with `--global`.

Explicit-only meta skill for running another agent skill repeatedly, defaulting to 3 iterations, with one final net report.

Interview the user relentlessly about a plan or design until reaching shared understanding, resolving each branch of the decision tree. Use when user wants to stress-test a plan, get grilled on their design, or mentions "grill me".

Use for debriefs or saved findings/reports from the current task. Creates `.ai/debriefs/<slug>/index.html`; `--md` writes `index.md` instead.

Use to find, compare, and recommend current tools, packages, libraries, CLIs, VSCode extensions, agent skills, databases, or infrastructure options for a development task.

Use for CoinGecko/cg CLI crypto market data: prices, market cap, trending coins, top gainers/losers, coin search, historical data, or OHLC.

Use to create interactive single-file HTML playgrounds/explorers/tools with visual controls, live preview, and prompt copy-out.

Use for Tailwind v4 styling: add/fix classes, configure or migrate Tailwind, use tailwind-variants, or tw-animate-css.

Use for Vitest in TypeScript React/Next.js: write, run, or debug unit/component tests, mocks, testing utilities, and coverage.

This skill should be used when the user asks to create a monochrome technical diagram, schematic, or systems/architecture diagram in black-and-white line-art style. Trigger phrases include "create a diagram", "monochrome diagram", "systems diagram", "draw a schematic".

This skill should be used when the user asks to consult Codex/GPT for planning or code review, or wants a second opinion on code. Trigger phrases include "ask Codex", "consult Codex", "second opinion on code", "consult the oracle". NOT for implementation tasks.

Use to squash a feature branch into one commit and rebase onto base: squash commits/branch, clean up commits, flatten branch history, prepare clean PR commit.

Run only on explicit invocation. Orchestrates end-to-end implementation, scales from direct work to subagents, and finishes with code-polish.