skills
AGENTS.md1 starsChanged 6 months agoMITView on GitHub
This repo is a private collection of personal agent skills. Keep skills/ as the canonical source tree so the same skill folders can be consumed by Codex, Claude Code, and the skills CLI. Each skill must have a SKILL.md with YAML frontmatter and Markdown instructions: Portable requirements: Codex can read direct skills and plugin-packaged skills. Useful commands: For a private GitHub repo, prefer SSH URLs if local SSH auth is already configured. Codex's skill installer can also use existing git…
- Reads credentials
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 31,843 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 repository28 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- add-mcp-serverSkill
- agents-md-initSkill
- agents-md-maintainSkill
- codexSkill
- coding-benchmarksSkill
- context7-docsSkill
- craft-mcpSkill
- deep-code-reviewSkill
- github-opsSkill
- herdr-omp-orchestrationSkill
- herdr-orchestrationSkill
- himalaya-emailSkill
- image-annotatorSkill
- linkup-searchSkill
- mise-manageSkill
- omniroute-opsSkill
- omp-opsSkill
- parallel-searchSkill
- playwriterSkill
- prompt-craftSkill
- session-retroSkill
- skill-createSkill
- skills-manageSkill
- surge-cliSkill
- tavily-searchSkill
- web-search-routerSkill
- workflow-auditSkill
The file
We split this AGENTS.md into the rules it gives an agent
8 rules. About 0 are common, 0 are distinctive, and 8 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
This author's own choices. Worth a look.
8
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 30
Keep SKILL.md focused. Move long reference material into files linked from SKILL.md.
- Very specific: line 31
Use optional frontmatter such as compatibility, metadata, and allowed-tools sparingly because agent support varies.
- Very specific: line 32
Prefer portable fields first.
- Very specific: line 33
Use references/ for long context, scripts/ for deterministic helpers, assets/ for reusable templates/static files, and agents/openai.yaml …
- Very specific: line 111
Commit and push skills/; remote installs only see committed files.
- Very specific: line 112
Keep root skills/ canonical.
- Very specific: line 113
Keep .codex-plugin/plugin.json, .claude-plugin/plugin.json, .agents/plugins/marketplace.json, and .claude-plugin/marketplace.json in sync …
- Very specific: line 114
Keep skills/<skill>/agents/openai.yaml prompts accurate when renaming skills or changing their trigger behavior.
≈ “Common” is an estimate: at least 1 in 5 of the 40,246 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
Personal Skills Repository
This repo is a private collection of personal agent skills. Keep skills/ as the canonical source tree so the same skill folders can be consumed by Codex, Claude Code, and the skills CLI.
Canonical Layout
skills/
<skill-name>/
SKILL.md
references/ # optional, detailed docs loaded on demand
scripts/ # optional, executable helpers
assets/ # optional, templates or static resources
agents/ # optional, platform UI/dependency metadataEach skill must have a SKILL.md with YAML frontmatter and Markdown instructions:
---
name: <kebab-case-name>
description: What the skill does and when to use it.
---Portable requirements:
nameis required for Codex and the Agent Skills spec. Use lowercase letters, numbers, and hyphens; keep it under 64 characters; avoid leading/trailing hyphens and repeated hyphens; match the parent directory name.descriptionis required, non-empty, and should stay under 1024 characters. Front-load trigger words because tools use the description for automatic skill selection.- Keep
SKILL.mdfocused. Move long reference material into files linked fromSKILL.md. - Use optional frontmatter such as
compatibility,metadata, andallowed-toolssparingly because agent support varies. - Prefer portable fields first. Platform-specific fields such as Claude Code
context,agent,disable-model-invocation, anduser-invocableshould be added only when the skill intentionally needs that runtime behavior. - Use
references/for long context,scripts/for deterministic helpers,assets/for reusable templates/static files, andagents/openai.yamlfor Codex-specific interface metadata.
Codex
Codex can read direct skills and plugin-packaged skills.
- Current Codex interface metadata lives at
skills/<skill>/agents/openai.yaml; keepdefault_promptexamples aligned with actual skill names. - Direct repo-scoped Codex skills normally live under
.agents/skills, but this repo keepsskills/as canonical and uses plugin/installer metadata for distribution. - Direct personal skills can be installed into user-level skill locations with
$skill-installeror theskillsCLI. - Codex plugin packaging uses
.codex-plugin/plugin.json; this repo's manifest points Codex at./skills/. - The repo-level Codex marketplace at
.agents/plugins/marketplace.jsonexposes this repo as a local/Git-backed private plugin catalog. - The Codex plugin namespace is
phntm. Namespacing applies to plugin-packaged skills, not to normal standalone installs through theskillsCLI.
Useful commands:
# List skills visible to the skills CLI from this repo
DISABLE_TELEMETRY=1 skills add . --list
# Install all skills globally into Codex and Claude Code via the skills CLI
DISABLE_TELEMETRY=1 skills add git@github.com:phntm7/skills.git --skill '*' -g -a codex -a claude-code -y
# Add this repo as a Codex plugin marketplace
codex plugin marketplace add git@github.com:phntm7/skills.gitFor a private GitHub repo, prefer SSH URLs if local SSH auth is already configured. Codex's skill installer can also use existing git credentials or GITHUB_TOKEN/GH_TOKEN for private repos.
Claude Code
Claude Code supports standalone skills and plugin-packaged skills.
- Personal standalone skills live in
~/.claude/skills/<skill>/SKILL.md. - Project standalone skills live in
.claude/skills/<skill>/SKILL.md. - Plugin skills live in
skills/<skill>/SKILL.mdinside a plugin root. Current skill names are unprefixed inSKILL.md; plugin invocation adds thephntmnamespace, for example/phntm:skill-create. - Claude plugin packaging uses
.claude-plugin/plugin.json; this repo's manifest treats the repo root as the plugin root. - The Claude marketplace file at
.claude-plugin/marketplace.jsonexposes this repo as a private marketplace with one plugin.
Useful commands:
# Test the plugin directly from a local clone
claude --plugin-dir .
# In Claude Code, add the private marketplace, then install the plugin
/plugin marketplace add git@github.com:phntm7/skills.git
/plugin install phntm@phntmskills CLI
The skills CLI already discovers the current repo shape because it scans skills/<skill>/SKILL.md.
Important behavior:
skills add . --listdetects local skills without installing them.--skill '*'installs all discovered skills.-a codex -a claude-codetargets Codex and Claude Code.--copycopies files instead of symlinking.--full-depthsearches nested skill trees more aggressively if a rootSKILL.mdever exists.- Standalone installs keep the skill's own
name, such asprompt-craft; they do not use thephntmplugin namespace.
Telemetry:
skillstelemetry is enabled by default for public installs and can include skill names, source, agents, and skill file paths.- In
skills@1.5.6, telemetry is disabled when eitherDISABLE_TELEMETRYorDO_NOT_TRACKis present in the environment. - There is no persistent
skills telemetry disablecommand in the current CLI help. To disable globally, export one of those variables from the shell environment, preferably in~/.zshenvif every zsh process should inherit it:
export DISABLE_TELEMETRY=1
# or use the broader console opt-out convention:
export DO_NOT_TRACK=1For one-off private repo operations, prefix commands with DISABLE_TELEMETRY=1.
Maintenance Checklist
- Commit and push
skills/; remote installs only see committed files. - Keep root
skills/canonical. Avoid duplicating the same skills under.agents/skillsor.claude/skillsin this repo unless a specific local workflow requires it. - Keep
.codex-plugin/plugin.json,.claude-plugin/plugin.json,.agents/plugins/marketplace.json, and.claude-plugin/marketplace.jsonin sync with the repo's skill collection and version. - Keep
skills/<skill>/agents/openai.yamlprompts accurate when renaming skills or changing their trigger behavior. - Bump plugin versions when publishing meaningful manifest or skill changes through plugin marketplaces. Convention: a separate
chore(release): bump plugin version to X.Y.Zcommit after the change commits; minor bump for adding or removing a skill, patch for fixes. Version fields live in.codex-plugin/plugin.json,.claude-plugin/plugin.json, and.claude-plugin/marketplace.json—.agents/plugins/marketplace.jsonhas no version field. - Validate discovery before publishing:
DISABLE_TELEMETRY=1 skills add . --list
for skill in skills/*; do python3 skills/skill-create/scripts/validate_skill.py "$skill"; done
python3 -m json.tool .codex-plugin/plugin.json >/dev/null
python3 -m json.tool .claude-plugin/plugin.json >/dev/null
python3 -m json.tool .agents/plugins/marketplace.json >/dev/null
python3 -m json.tool .claude-plugin/marketplace.json >/dev/nullSources
- Agent Skills Specification: https://agentskills.io/specification
- Codex skills: https://developers.openai.com/codex/skills/
- Codex plugins: https://developers.openai.com/codex/plugins/build
- Claude Code skills: https://code.claude.com/docs/en/skills
- Claude Code plugins: https://code.claude.com/docs/en/plugins
- Claude Code plugin reference: https://code.claude.com/docs/en/plugins-reference
- Claude Code plugin marketplaces: https://code.claude.com/docs/en/plugin-marketplaces
- skills CLI docs: https://www.skills.sh/docs/cli
- skills CLI repository: https://github.com/vercel-labs/skills
- Console
DO_NOT_TRACKconvention: https://donottrack.sh/
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.

