agent-skills
CLAUDE.md4 starsChanged 8 months agoCC0-1.0View on GitHub
Agent Skills (v5.0.0) is a portable library of Markdown-based skills that encode professional software development practices for any AI coding agent. Skills are harness-agnostic — they work on Claude Code, Codex, Cursor, Windsurf, OpenCode, Goose, Amp, Aider, and any Agent Skills-compatible harness. This is a content-only repository: no build system, no test runner, no compiled artifacts. The deliverables are SKILL.md files and their reference documentation, installed via npx skills add jwilger/agent-skills. There are no build, lint, or test commands. All…
We split this CLAUDE.md into the rules it gives an agent
6 rules. About 0 are common, 1 is distinctive, and 5 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.
5
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 22
Commit and PR hygiene: Never add AI co-author lines (Co-Authored-By), "Generated with", "Created by", or any similar taglines to commit …
- Very specific: line 45
Skills are the single source of truth: All practices live in SKILL.md files.
- Distinctive: line 64
When creating or modifying a skill:
- Very specific: line 66
Follow the template at skills/.template/SKILL.md.example
- Very specific: line 67
Use the frontmatter reference at skills/.template/FRONTMATTER.md
- Very specific: line 70
Keep SKILL.md under 500 lines.
≈ “Common” is an estimate: at least 1 in 5 of the 25,680 CLAUDE.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
CLAUDE.md
This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
Project Overview
Agent Skills (v5.0.0) is a portable library of Markdown-based skills that encode professional software development practices for any AI coding agent. Skills are harness-agnostic — they work on Claude Code, Codex, Cursor, Windsurf, OpenCode, Goose, Amp, Aider, and any Agent Skills-compatible harness.
This is a content-only repository: no build system, no test runner, no compiled artifacts. The deliverables are SKILL.md files and their reference documentation, installed via npx skills add jwilger/agent-skills.
Development Workflow
There are no build, lint, or test commands. All changes are to Markdown files. Quality is maintained through:
- Adherence to the skill template (
skills/.template/SKILL.md.example) - Token budget compliance (see below)
- Frontmatter correctness (
skills/.template/FRONTMATTER.md) - Section order enforcement (Value → Purpose → Practices → Enforcement Note → Verification → Dependencies)
Version is tracked in the VERSION file (semver). Use /bump to increment.
Commit and PR hygiene: Never add AI co-author lines (Co-Authored-By), "Generated with", "Created by", or any similar taglines to commit messages or PR descriptions. The use of AI tooling should not be called out in version control metadata.
Architecture
Skill Tiers
| Tier | Skills | Token Budget |
| 0 (Bootstrap) | bootstrap | 1000 |
| 1 (Core Process) | tdd, domain-modeling, code-review, architecture-decisions, event-modeling, ticket-triage, design-system | 3000 |
| 2 (Team Workflows) | ensemble-team, task-management | 4000 |
| 3 (Utility) | debugging-protocol, user-input-protocol, memory-protocol, agent-coordination, session-reflection | 3000 |
| 4 (Factory Pipeline) | pipeline, ci-integration, factory-review | 3000 |
| Advanced | mutation-testing, atomic-design | 3000 |
Skill Directory Structure
Each skill lives in skills/<name>/ and contains:
SKILL.md— the skill itself (frontmatter + six required body sections)references/— detailed supplementary docs loaded on demand by the agent
Key Design Principles
- Skills are the single source of truth. All practices live in SKILL.md files. No separate plugin layer.
- Enforcement proportional to capability. Skills adapt: structural enforcement (context isolation, handoff schemas) on capable harnesses, advisory guidance on simpler ones. Optional hook templates in
skills/tdd/references/hooks/add mechanical enforcement on Claude Code. - Graceful degradation. Missing dependencies recommend installation but never block execution. Every skill provides value standalone.
- Dependencies form a DAG. Declared in frontmatter
metadata.requires. No circular dependencies. Skills reference each other by name, never internal structure. - Progressive disclosure. SKILL.md stays concise (within token budgets). Extended examples and detailed guides live in
references/.
Factory Pipeline (Tier 4)
The optional pipeline orchestrates autonomous build-and-ship: Human plans → Agent builds (TDD pairs, code review, mutation testing, CI) → Human reviews. Three autonomy levels (conservative/standard/full). Pipeline state lives in .factory/ in consumer projects. At full autonomy on Claude Code, parallel slices run in isolated git worktrees.
TDD Execution Strategies
The tdd skill auto-detects harness capabilities and selects:
- Subagents — isolated phases via Agent tool, with named personas when
.claude/agents/definitions exist (Claude Code, Codex) - Chaining — sequential roles in single context (all harnesses)
- Guided — manual phase control via
/tdd red,/tdd green, etc.
Authoring Skills
When creating or modifying a skill:
- Follow the template at
skills/.template/SKILL.md.example - Use the frontmatter reference at
skills/.template/FRONTMATTER.md - Maintain all six body sections in order: Value, Purpose, Practices, Enforcement Note, Verification, Dependencies
- Stay within the tier's token budget
- Keep SKILL.md under 500 lines; move detail to
references/ - Enforcement Note must honestly state guarantees at each level (advisory/structural/mechanical)
- Verification items must be binary and observable
- Shell fragments in SKILL.md: read-only environment detection only — no writes, no network calls, no package installation
Other versions1 project uses a version of this.
Word for word · 1
- jwilger/agent-skills-archive-20260705★ 3100% same
Edited · 0
None found.
Wrote their own version · 0
None found.
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository21 other files it gives its agents.
- bumpSkill
- agent-coordinationSkill
- architecture-decisionsSkill
- atomic-designSkill
- bootstrapSkill
- ci-integrationSkill
- code-reviewSkill
- debugging-protocolSkill
- design-systemSkill
- domain-modelingSkill
- ensemble-teamSkill
- event-modelingSkill
- factory-reviewSkill
- memory-protocolSkill
- mutation-testingSkill
- pipelineSkill
- session-reflectionSkill
- task-managementSkill
- tddSkill
- ticket-triageSkill
- user-input-protocolSkill
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
Reviews can't be read right now.
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.