ywc-agent-toolkit
AGENTS.md8 starsChanged 5 months agoMITView on GitHub
This repository distributes ywc-* skills for Claude Code and Codex. Claude Code skills live in claude-code/skills/<skill-name>/; Codex skills live in codex/skills/<skill-name>/. Each skill requires SKILL.md plus Tier 1 localized README files: README.md, README.en.md, README.ja.md, and README.ko.md. Generated Tier 2 READMEs, README.zh.md and README.es.md, are maintained for skills that ship Chinese and Spanish documentation. Codex skills also require agents/openai.yaml. Claude Code custom agents live in claude-code/agents/<agent-name>.md (one Markdown file per agent). Codex custom agents live in codex/agents/<agent-name>.toml (one TOML per agent).…
Where files are headingNew AGENTS.md files have more rules about keeping secrets out of the code, and fewer about code style.
Of the AGENTS.md files we read, the 33,315 first seen in the last six months, against the older ones: the kind of rule that gained most share, and the kind that lost most. About files like this in general, not this one, and an estimate, since a kind is a pattern in the words.
In this repository4 other files it gives its agents.
- codex/AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- graftSkill
- ywc-toolkit-evalSkill
The file
We split this AGENTS.md into the rules it gives an agent
7 rules. About 0 are common, 1 is distinctive, and 6 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
6
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- could be automatic: a setting could enforce it
- Very specific: line 28
Follow Conventional Commit style used in history: feat:, fix:, docs:, i18n:, ci:, or chore:.
- Very specific: line 32
Do not commit secrets or local install artifacts.
- Very specific: line 44
Run graft map first — a token-budgeted orientation (dir clusters, hubs, hotspots), no LLM, no key.
- Very specific: line 47
Run graft ask "<your question>" --source → ranked nodes with the relevant code spans inlined (each hit's ≤8-line crux by default.
- Very specific: line 59
Add --direction out for what it calls, or --depth N to walk transitively for the full blast radius.
- Distinctive: line 67
If a returned span is truncated ("+N more lines"), open the file at that exact range before finalizing.
- Very specific: line 72
After big code changes, refresh the graph with graft build (deterministic, no API key, $0).
≈ “Common” is an estimate: at least 1 in 5 of the 42,283 AGENTS.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Repository Guidelines
Project Structure & Module Organization
This repository distributes ywc-* skills for Claude Code and Codex. Claude Code skills live in claude-code/skills/<skill-name>/; Codex skills live in codex/skills/<skill-name>/. Each skill requires SKILL.md plus Tier 1 localized README files: README.md, README.en.md, README.ja.md, and README.ko.md. Generated Tier 2 READMEs, README.zh.md and README.es.md, are maintained for skills that ship Chinese and Spanish documentation. Codex skills also require agents/openai.yaml. Claude Code custom agents live in claude-code/agents/<agent-name>.md (one Markdown file per agent). Codex custom agents live in codex/agents/<agent-name>.toml (one TOML per agent). Shared Codex material belongs in codex/skills/references/ or codex/skills/scripts/. Root files such as README.md, CONTRIBUTING.md, CHANGELOG.md, VERSION, plugin.json, and translations.json describe distribution metadata and release state.
Build, Test, and Development Commands
bash scripts/install.sh --list: list all installable skills and verify the install script can scan the repository.bash scripts/install.sh --cc ywc-plan: install one Claude Code skill into${CLAUDE_SKILLS_DIR:-~/.claude/skills}for local testing.bash scripts/install.sh --codex ywc-plan: install one Codex skill into${CODEX_HOME:-~/.codex}/skills.bash scripts/install.sh --cc-agents: install Claude Code custom agents into${CLAUDE_AGENTS_DIR:-~/.claude/agents}.bash scripts/install.sh --codex-agents: install Codex custom agents into${CODEX_HOME:-~/.codex}/agents.bash scripts/install.sh --all: install both Claude Code and Codex bundles (skills + agents).bash scripts/validate.sh: run the local validation mirror for CI structure checks.bash scripts/translate.sh --dry-run: preview Tier 2 translation regeneration without API calls.
Coding Style & Naming Conventions
Use portable Bash for scripts with set -euo pipefail. Keep skill names and directories in kebab case, for example ywc-tech-research. SKILL.md frontmatter must include name: and description:. Codex SKILL.md frontmatter must contain only those two fields; do not copy Claude-only fields such as version, category, or requires. Keep Markdown concise, instructional, and example-driven.
Testing Guidelines
There is no package test runner. Treat bash scripts/validate.sh as the required pre-PR test. For install changes, also run targeted install/list commands. New or changed skills should be checked for required Tier 1 README locale files, maintained Tier 2 README.zh.md / README.es.md files when present, matching name: frontmatter, and Codex agents/openai.yaml metadata.
Commit & Pull Request Guidelines
Follow Conventional Commit style used in history: feat:, fix:, docs:, i18n:, ci:, or chore:. Example: docs: update skill authoring guide. PRs should explain the change, link the related issue when available, and pass CI checks for validation, shellcheck, markdownlint, and translation warnings.
Security & Configuration Tips
Do not commit secrets or local install artifacts. Use CLAUDE_SKILLS_DIR and CODEX_HOME to test installs in temporary directories before touching real user skill folders.
<!-- graft:start -->Graft — repo context graph
This repo is indexed in graft/: small linked markdown nodes that explain each
system and carry exact file:line spans, kept in sync with the code through git.
For ANY task here — understanding how something works, finding where code lives,
or scoping a change — get context from the graph before grepping or opening
source files. Re-ask freely (it's cheap) and reuse literal identifiers you
already have (symbol, error string, file name) as the query. New to this repo?
Run graft map first — a token-budgeted orientation (dir clusters, hubs,
hotspots), no LLM, no key.
- Run
graft ask "<your question>" --source→ ranked nodes with the relevant code spans inlined (each hit's ≤8-line crux by default;--fullfor whole definitions when the crux isn't enough). Match the tool to the task shape: for understanding or editing, the top node IS the answer — cite itscovers:file:line spans and edit straight from--source. For exhaustive tasks ("every occurrence / every caller of this pattern"), ranked results are top-N, not complete — rungraft grep "<literal>"instead (exhaustive over indexed files, grouped by enclosing symbol), falling back to rawgrep -rnonly for unindexed files. graft skeleton <file>→ every definition's signature + span, ~10× cheaper than reading the file; use it to skim an API surface.graft callers <symbol>gives precomputed, exact edges — who calls this. Add--direction outfor what it calls, or--depth Nto walk transitively for the full blast radius. For structural questions, skip ranking and use this directly.- Or browse:
graft/INDEX.mdlists every node; follow the links. - Monorepos and folders of multiple repos rank fairly across sub-projects —
hits carry
[scope/]labels naming which one they're from. Narrow withgraft ask "<task>" --in <scope>/once you know where you're working.
If a returned span is truncated ("+N more lines"), open the file at that exact range before finalizing. Only open source files when a node genuinely lacks a needed detail, and then at the exact file:line the node points to — never re-read whole files.
After big code changes, refresh the graph with graft build (deterministic,
no API key, $0).
<!-- graft:end -->Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

