https://github.com/ribeirogab/agent-skills
Reusable agent skills for any tool that supports the open agent skills standard.
https://github.com/ribeirogab/agent-skills
agent-skills ai-agents anthropic claude-code claude-skills harness skills
Last synced: 14 days ago
JSON representation
Reusable agent skills for any tool that supports the open agent skills standard.
- Host: GitHub
- URL: https://github.com/ribeirogab/agent-skills
- Owner: ribeirogab
- License: mit
- Created: 2026-04-30T20:10:14.000Z (about 2 months ago)
- Default Branch: main
- Last Pushed: 2026-05-15T18:01:54.000Z (about 1 month ago)
- Last Synced: 2026-05-15T19:55:28.219Z (about 1 month ago)
- Topics: agent-skills, ai-agents, anthropic, claude-code, claude-skills, harness, skills
- Language: Python
- Size: 570 KB
- Stars: 0
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- Contributing: CONTRIBUTING.md
- License: LICENSE
- Code of conduct: CODE_OF_CONDUCT.md
- Security: SECURITY.md
- Notice: NOTICE.md
- Agents: AGENTS.md
Awesome Lists containing this project
README
# agent-skills
Reusable agent skills for any tool that supports the open agent skills standard — Claude Code, Codex, Cursor, OpenCode, Gemini CLI, Aider, Cline, Augment, and others. Skills are framework-agnostic by design: a skill is a folder with a `SKILL.md` and any helpers it needs; agents discover them by description.
> Personal project, solo maintenance, best-effort, no SLA. Published so anyone can pull a skill into their setup.
---
## Skills
### `memex`
Idempotently scaffolds a memex — an externalized, navigable project memory for agents — into any repository: a `.vault/` knowledge vault, an `AGENTS.md` (with a `CLAUDE.md` symlink for back-compat), spec/plan/task templates, plus a set of bundled `memex-*` slash commands (open-pr, learn, spec, review-spec, sweep) and companion skills (brainstorming, recall, writing-plans). Audit-first, autonomous-fix, with a Phase-5 validator. Safe to re-run.
**Install:**
```bash
npx skills add ribeirogab/agent-skills --skill memex
```
**Use:** point an agent at any repo where you want the memex installed.
> "Audit the memex in this repo and scaffold whatever is missing."
After the first run the repo has a working `.vault/` vault, the `memex-*` commands, and the companion skills, all dogfood-tested by the memex's own 13-check validator.
**Source:** [`skills/memex/SKILL.md`](skills/memex/SKILL.md)
---
### `skill-improver`
Audits an existing agent skill against the canonical authoring rules (frontmatter, layout, body style, progressive disclosure, description quality, degrees of freedom) and applies safe fixes autonomously. Defers high-regression-risk findings for manual review. Self-contained — bundles vendored copies of `quick_validate.py` and `package_skill.py` so it does not require the upstream `skill-creator` to be installed.
**Install:**
```bash
npx skills add ribeirogab/agent-skills --skill skill-improver
```
**Use:** point it at any skill folder.
> "Audit the skill at `skills/my-skill` and apply safe fixes."
The skill walks a 10-section canonical checklist, applies anything `Low` or `Medium` regression-risk autonomously, and produces a final report with a `Skipped` section for `High`-risk findings the maintainer should review by hand.
**Source:** [`skills/skill-improver/SKILL.md`](skills/skill-improver/SKILL.md)
---
## Repository layout
What matters for users is just `skills/`. Each subfolder is one publishable skill:
```
agent-skills/
├── skills/
│ ├── memex/
│ └── skill-improver/
├── LICENSE # MIT, this repository's original work
├── NOTICE.md # attribution for any vendored content inside skills/
├── CONTRIBUTING.md
├── CODE_OF_CONDUCT.md
├── SECURITY.md
└── README.md
```
The repository also contains `.agents/`, `.claude/`, `.vault/`, and `evals/` — local-only dirs used by the maintainer to dogfood the memex, run skill evaluations, and store personal project notes. They are not part of the published surface and are not what `npx skills add` installs.
## License
This repository's original work is licensed under the [MIT License](LICENSE). Any vendored third-party content inside the published skills is governed by its upstream license; see [`NOTICE.md`](NOTICE.md) for the full attribution.
## Contributing
Pull requests welcome — see [`CONTRIBUTING.md`](CONTRIBUTING.md) for scope, the quality bar, and the per-PR checklist. By participating, you agree to the [Code of Conduct](CODE_OF_CONDUCT.md). Security concerns go to [`SECURITY.md`](SECURITY.md).