agent-skills
Cross-agent skills and Pi extensions
Open source Open in the app JSON README (API)
About
Cross-agent skills and Pi extensions
Details
- Kind
- Plugins
- Topic
- No topic detected
- Publisher
- sheurich
- Origin
- gemini
- Category
- ferramentas
- Version
- 1.0.0
- Open pull requests
- 1
- Last push
- 2026-07-10T13:26:25Z
- Repository state
- ativo
- Language
- TypeScript
- License
- MIT
- Added
- 2026-08-30 14:13:39
- Updated
- 2026-08-30 14:13:39
- Origin id
sheurich/agent-skills
README
# Agent Skills [Agent skills](https://agentskills.io) that work across coding agents. ## Install | Agent | Command | | --- | --- | | Pi | `pi install https://github.com/sheurich/agent-skills` | | Claude Code | `/plugin marketplace add sheurich/agent-skills` | | Gemini CLI | `gemini extensions install sheurich/agent-skills` | | npx skills | `npx skills add sheurich/agent-skills --all -g` | Install a single skill with npx: ```bash npx skills add sheurich/agent-skills -s <skill-name> -g ``` ## Skills | Skill | Description | | --- | --- | | [readiness-review](skills/readiness-review/SKILL.md) | Structured readiness assessment before proceeding to the next environment or phase. | | [semgrep-scan](skills/semgrep-scan/SKILL.md) | Scan changed files with Semgrep for security and correctness issues before committing. | | [skill-authoring](skills/skill-authoring/SKILL.md) | Write and refine agent skills. Use when generating skills from docs, extracting from sessions, designing tool interfaces, managing lifecycle, or writing for dual-harness patterns. | | [swival](skills/swival/SKILL.md) | Delegate tasks to Swival for self-reviewed code changes, secret-safe operations, cached analysis, and A2A agent serving. | ## Plugins | Plugin | Description | | --- | --- | | [auto-summarize](plugins/auto-summarize/) | Rolling session summary and automatic session naming for Pi. | ## Pi Packages | Package | Description | | --- | --- | | [swival-subagent](packages/swival-subagent/) | Pi extension that delegates tasks to a swival subprocess (reviewer loop, AgentFS sandbox, secret encryption). | ## Contributing See [CONTRIBUTING.md](CONTRIBUTING.md) for how to add skills, write tests, and run validation. ## Repo Structure ```text skills/ # Simple skills (SKILL.md only) plugins/ # Claude Code plugins (own manifests, hooks, LSP, MCP) packages/ # Pi packages (pi.extensions in package.json) tests/scenarios/ # One scenario per skill or package scripts/validate-extra.sh # Repo-specific validation checks vendor/agent-validate/ # Vendored validator validate.sh # Thin wrapper → vendor/agent-validate .claude-plugin/marketplace.json gemini-extension.json ``` ## Validation ```bash ./validate.sh # run all checks ./validate.sh --skip json # skip specific checks ``` CI runs [agent-validate](https://github.com/sheurich/agent-validate) on every push and pull request. ## License [MIT](LICENSE)