Reusable agent skills collected in one repo.
This repository is the source for the skills under skills/. It is intentionally lightweight: the repo mainly contains skill folders with SKILL.md instructions plus any supporting scripts, references, or templates each skill needs.
- Custom and adapted skills for coding, ops, writing, review, debugging, and workflow automation
- Supporting assets that live next to a skill, such as Python helpers, shell scripts, templates, and reference notes
- A local install/source lock file in
.skill-lock.jsonshowing where some imported skills came from
.
├── skills/
│ ├── <skill-name>/
│ │ ├── SKILL.md
│ │ ├── scripts/
│ │ ├── references/
│ │ └── templates/
└── .skill-lock.json
Each skill is self-contained. The contract is simple: if a tool or agent knows how to discover SKILL.md files, it can use this repo as a skill source.
skills/retro provides a retrospective workflow for reviewing recent coding sessions and improving skills, agents, and workflows. Its helper script, skills/retro/retro_extract.py, currently parses Claude Code-style .jsonl session transcripts from ~/.claude/projects.
That is narrower than Angie’s version in angie/agentic-setup, which has expanded retro_extract.py support for multiple backends and a vendoring/install layer around it. This repo is the upstream skill source; Angie’s repo is a broader personal agent environment that consumes repos like this one.
How you install these skills depends on the agent framework you use, but the common pattern is:
- Clone this repo somewhere your agent tooling can read.
- Point your tool at the
skills/directory, or copy/link individual skill folders into its skill search path. - Keep third-party provenance in sync if you import or update external skills.
This repo does not currently include a universal installer or manifest of its own.
angie/agentic-setup: a full personal agent environment with vendoring, manifest-driven sync, and personal overrides; it explicitly vendorsdamacus/agent-skillsas an opt-in upstream forretrocitypaul/.dotfiles: a major upstream in Angie’s setup forCLAUDE.md, skills, commands, and agentsmintuz/claude-plugins: another upstream in Angie’s setup for skills, commands, and agents
Several skills in this repo were imported or adapted from other public repositories. The current local lockfile references these sources:
anthropics/skillsvercel-labs/skillsmicrosoft/playwright-cliaddyosmani/web-quality-skillsodyssey4me/agent-skillsnicolaischmid/agent-skillsmarkpitt/claude-skillsjeffallan/claude-skills
If you want stricter attribution, the next useful step would be adding per-skill provenance metadata or a small generated table from .skill-lock.json.