⚠️ Archived — moved into the delego plugin. These skills now live in Delego-Dev/plugin (the single source of truth). Install everything in one command:

/plugin marketplace add Delego-Dev/plugin
/plugin install delego@delego

Or à la carte: clone the plugin repo and run its install.sh. The content below is a frozen snapshot.

delego skills

Claude Code skills for working with delego — the policy & audit firewall for agent actions. Skills are the interactive surface: you drive them in a Claude Code session to set delego up, write a policy, triage approvals, and read the audit trail.

Quick start

# in your project: clone the skills into the folder Claude Code reads
git clone https://github.com/Delego-Dev/skills .claude/skills

Skills hot-load (no restart). Then, in Claude Code, run /delego-init — it installs delego, creates the firewall home, wires the MCP server, and verifies. That's the whole setup.

Skill	Invoke	Use it to…
delego-init	/delego-init	Install delego, create a project-scoped firewall home + keys, wire the MCP server into .mcp.json, and verify.
delego-policy-drafter	/delego-policy-drafter	Draft or harden a policy.yaml from what your agent should/shouldn't do — schema-correct, fail-closed, footguns covered.
delego-approval-triage	/delego-approval-triage	Review pending approvals, sanity-check each against intent, and approve/deny.
delego-audit-explainer	/delego-audit-explainer	Verify the signed audit chain and explain what the agent did and why.

For autonomous review (policy/broker reviewers, audit investigator) see the companion agents repo.

Setup (Claude Code)

Claude Code discovers skills one level deep at .claude/skills/<name>/SKILL.md. The skill directories live at this repo's root, so the simplest install is to clone it straight into .claude/skills/:

cd your-project
git clone https://github.com/Delego-Dev/skills .claude/skills

Skills hot-load — available immediately, no restart. Update with git -C .claude/skills pull. (The repo's README, scripts/, etc. sit alongside the skill dirs and are ignored by skill discovery.)

Already have a .claude/skills/? A nested clone won't be found (skills are one level deep), so keep this repo separate and symlink each skill in:

git clone https://github.com/Delego-Dev/skills .claude/delego-skills
.claude/delego-skills/install.sh        # symlinks each skill into .claude/skills/

Invoke

Type the slash command, optionally with a hint:

/delego-init
/delego-policy-drafter   our agent reads GitHub issues and posts comments; nothing else
/delego-approval-triage
/delego-audit-explainer   why was the transfer denied?

Claude will also invoke a skill on its own when your request matches its description — e.g. "set up delego in this repo" triggers delego-init.

Typical flow

1. /delego-init — stand up delego + the MCP server in this project.
2. /delego-policy-drafter — write the policy for your agent.
3. Review it with the delego-policy-reviewer agent (from the agents repo).
4. At runtime: /delego-approval-triage to decide parked actions, and /delego-audit-explainer to read the signed trail.

Contributing

Each skill is a directory with a SKILL.md (YAML frontmatter: kebab-case name matching the directory, plus a description that says when to use it). python scripts/validate.py (run in CI) checks every skill. PRs welcome — fork and open one.

Roadmap

Planned skills: a broker-adapter scaffolder (generate a BrokerAdapter for your vault/proxy), a policy-from-traffic drafter (infer a starting policy from request logs), and a deny-explainer ("why was X denied, and what's the safe policy change?").

License

Apache-2.0. Built for delego; see the wire specification.