Skill Steward

Meta skills for the Agent Skills ecosystem — validate, govern, and document portable SKILL.md packages. Not a domain skill catalog (React, Flutter, cloud recipes live elsewhere). Not a skill installer (Skillkit and skills.sh cover distribution).

Install on Cursor, Claude Code, Codex, Windsurf, GitHub Copilot, and 15+ tools via npx skills.

Charter: docs/NORTH_STAR.md · Docs: docs.page/arenukvern/skill_steward · docs.json
Why / how: DESIGN_FAQ.md · DX_FAQ.md · Decisions · AGENTS.md (agent map)

skills.sh

How this fits my other work

Project	Role
Product Harnesses	Closed-loop tooling so agents can inspect and interact with running apps (CLI + agent integrations).
Skill Steward (this repo)	Meta layer — skills that help teams manage other skills: validation, ADRs, FAQ-driven docs, plan hygiene, harness culture.
Principles at work	Why — ethical AI boundaries, care for end users and builders, prototyping with feedback, artisan credit.

Same thread: useful docs for humans and agents, mechanical gates, and work worth people’s time. See FAQ-driven development for the documentation shape this repo dogfoods.

How the name was chosen: public product name and GitHub repo are Skill Steward (ADR 0008).

Install

All meta-skills:

npx skills add arenukvern/skill_steward

One skill:

npx skills add arenukvern/skill_steward --skill create-skill

Target specific agents:

npx skills add arenukvern/skill_steward -a cursor -a claude-code -y

Global install:

npx skills add arenukvern/skill_steward -g

Discover on skills.sh or:

npx skills find steward

Updating installed skills

Skills install as copies or symlinks under agent directories (for example .cursor/skills/ or .agents/skills/). When Skill Steward changes on GitHub, refresh your install with the skills CLI:

# Update every Skill Steward skill you have installed (project scope)
npx skills update -y

# Update only global installs
npx skills update -g -y

# Update only project-scoped installs
npx skills update -p -y

# Update one skill by name (as shown in `npx skills list`)
npx skills update north-star-governance -y

Re-install from GitHub when you want a clean pull of the whole marketplace or a single skill:

# Refresh all meta-skills from main
npx skills add arenukvern/skill_steward -y

# Refresh one skill
npx skills add arenukvern/skill_steward --skill harness-engineering-culture -y

# Same, but only for Cursor in this repo
npx skills add arenukvern/skill_steward -a cursor -y

See what is installed before updating:

npx skills list
npx skills list -g

Note: npx skills update tracks the source you installed from (GitHub main by default). It does not run Skill Steward’s pnpm run validate—that is for contributors. Hooks under plugins/ are separate; see plugins/README.md.

More commands: DX_FAQ.md (section Updating installed skills).

What belongs here

Meta and process capabilities only — inclusion criteria. Domain skills live in other repositories.

Available skills

Skill	Description
create-skill	Scaffold a new skill that passes validation and works with npx skills.
skill-spec-review	Audit a skill package against the Agent Skills spec.
plugin-marketplace-setup	Public/private skill & plugin marketplaces for Cursor, Claude, Codex, and npx skills.
skill-source-citations	Cite and persist URLs in references/sources.md when authoring skills.
skill-eval-improve	Tiered evals—rule-based pnpm run eval, Chrome/SkillOpt patterns, plugin-eval, human prompt suites.
adr-records	Write and maintain ADRs per adr.github.io.
faq-driven-docs	Maintain DESIGN_FAQ (why) and DX_FAQ (how) per FAQ-driven development.
concept-doc-store	Vectorless doc lattice (router, ADRs, concepts)—link to code for behavior; layered-docs style.
repo-brand-identity	Establish, maintain, and govern a repository's brand identity, custom/Shields.io status badges, palette, and tone.
ethical-stewardship	Establish, audit, and maintain core repository ethics as moral principles, constraints, and actionable rules.
multi-agent-handoff	Plan handoffs between specialized agents.
harness-engineering-culture	Agent-first harness design—CLI/MCP, mechanical gates, docs map (OpenAI harness engineering).
release-changelog-harness	Release/changelog tooling plus binary distribution contract (install.sh, GitHub Releases) for product harness repos.
mcp-harness-repo-maintainer	Maintain MCP/harness repos (product MCPs, libraries, CLI harnesses)—archetypes, contract gates, sibling layout.
north-star-governance	North Star charter, AGENTS.md map, plan hygiene (any format), docs.page wiring.

Standards

• Format: Agent Skills specification (SKILL.md + optional scripts/, references/, assets/)
• Registry: skills.sh indexes public repos with valid skills
• CLI: vercel-labs/skills (npx skills)

See docs/STANDARDS.md for the checklist used in this repo.

Repository layout

skill_steward/              # GitHub: Arenukvern/skill_steward
├── DESIGN_FAQ.md           # Why (decisions, charter)
├── DX_FAQ.md                 # How (install, validate, contribute)
├── docs/decisions/           # ADRs (incl. product name ADR 0008)
├── skills/                   # Meta-skills only
├── packages/steward_cli/       # Dart `steward` CLI — validate, list
├── plugins/
│   └── steward-validate-on-save/
├── templates/
│   ├── skill/
│   └── plugin/
├── skills.sh.json
├── CHANGELOG.md              # Generated via Changesets (ADR 0009)
├── .changeset/               # PR-time release notes
├── AGENTS.md
└── scripts/validate-skills.mjs

Contributing

1. Read AGENTS.md and CONTRIBUTING.md.
2. Add a skill under skills/{kebab-case-name}/ with SKILL.md.
3. Run pnpm run validate.
4. Update skills.sh.json and the skill table in this README.
5. Open a PR.

Validate locally

pnpm install
pnpm run steward:validate

Dart CLI directly (ADR 0006 / 0007):

cd packages/steward_cli && dart pub get && dart run :steward validate

Cursor validates skills/**/SKILL.md on save via [.cursor/hooks.json](.cursor/hooks.json).

License

MIT — see LICENSE.