agent-skills
CLAUDE.md25 starsChanged 9 months agoMITView on GitHub
This repo contains Claude Code skills for PHP development workflows. Skills are packaged as .skill files for distribution. Use the init_skill.py script from the skill-creator skill to scaffold a new skill: This creates a <skill-name>/ directory with a SKILL.md template and example resource directories. Every SKILL.md requires YAML frontmatter with name and description: Delete any example files from init_skill.py that aren't needed. When asked to review or improve a skill, read all files in the skill directory before making any…
In this repository11 other files it gives its agents.
- composer-upgradeSkill
- gritqlSkill
- laravel-advanced-conceptsSkill
- npm-upgradeSkill
- packagistSkill
- phpstan-developerSkill
- php-type-safetySkill
- rector-developerSkill
- remotion-3d-background-developerSkill
- serpapiSkill
- slidewire-developerSkill
The file
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 23
Use the init_skill.py script from the skill-creator skill to scaffold a new skill:
- Very specific: line 33
Every SKILL.md requires YAML frontmatter with name and description:
- Very specific: line 47
Keep the body under 500 lines.
- Very specific: line 56
Delete any example files from init_skill.py that aren't needed.
- Very specific: line 60
When asked to review or improve a skill, read all files in the skill directory before making any changes: SKILL.md and every file in …
- Distinctive: line 68
When trimming, preserve the key decision rules and non-obvious patterns.
≈ “Common” is an estimate: at least 1 in 5 of the 28,029 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
Agent Skills — Development Guide
This repo contains Claude Code skills for PHP development workflows. Skills are packaged as .skill files for distribution.
Project Structure
agent-skills/
├── <skill-name>/ # Skill source directory
│ ├── SKILL.md # Required: frontmatter + instructions
│ ├── references/ # Optional: reference docs loaded as needed
│ ├── scripts/ # Optional: executable scripts
│ └── assets/ # Optional: templates, images, etc.
├── <skill-name>.skill # Packaged distributable (zip with .skill extension)
├── package-all.sh # Packages all skills in the repo
└── README.mdDeveloping a Skill
1. Initialize
Use the init_skill.py script from the skill-creator skill to scaffold a new skill:
python3 ~/.claude/skills/skill-creator/scripts/init_skill.py <skill-name> --path .This creates a <skill-name>/ directory with a SKILL.md template and example resource directories.
2. Edit SKILL.md
Every SKILL.md requires YAML frontmatter with name and description:
---
name: my-skill
description: What this skill does and when it should trigger. Include trigger phrases and use cases here.
---
# Skill Title
Instructions for Claude...- The
descriptionis the primary trigger mechanism — be specific about when the skill should activate - Keep the body under 500 lines; move detailed content to
references/files - Reference files from
SKILL.mdwith clear guidance on when to read them
3. Add Resources
references/— Documentation Claude reads when needed (schemas, API docs, patterns)scripts/— Deterministic scripts for repeated operations; test them before packagingassets/— Templates and files used in output (not loaded into context)
Delete any example files from init_skill.py that aren't needed.
Reviewing an Existing Skill
When asked to review or improve a skill, read all files in the skill directory before making any changes: SKILL.md and every file in references/, scripts/, and assets/. Then look for:
- Duplication between SKILL.md and reference files — content that appears in both should live only in the reference file, with SKILL.md reduced to a one- or two-line pointer
- Duplication across reference files — consolidate or add cross-references
- Over-explained content — remove prose that restates what a code example already shows clearly; trust that Claude can read code
- Long sections that belong in references — any section in SKILL.md approaching 20+ lines on a single sub-topic is a candidate to move to a dedicated reference file
- Sections with no cross-reference — every reference file should be mentioned from SKILL.md with guidance on when to read it; orphaned reference files won't get used
When trimming, preserve the key decision rules and non-obvious patterns. What gets cut is repetition, verbose explanation of self-evident code, and content that is already comprehensively covered elsewhere.
Packaging a Skill
Package a single skill:
python3 ~/.claude/skills/skill-creator/scripts/package_skill.py <skill-name>/Package all skills in the repo:
./package-all.shThe packaging script validates the skill first (frontmatter, naming, structure) then creates <skill-name>.skill in the project root. Fix any validation errors and re-run.
Key Principles
- Concise is key — only include context Claude doesn't already have
- Progressive disclosure — metadata triggers, SKILL.md body loads on trigger, references load as needed
- No extra docs — don't create README, CHANGELOG, or other auxiliary files inside a skill directory; the skill should only contain what an AI agent needs to do the job
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.

