cli-setup skill

The cli-setup skill walks a user through configuring core-agent for interactive use. It’s bundled in SKILLS/cli-setup/ at the repo root.

What it covers

The four customization layers, in order:

1. Provider + model — .agents/config.json
2. Personality — AGENTS.md
3. Skills — .agents/skills/<name>/SKILL.md
4. Tools — built-ins + MCP servers

Plus permission posture (ask / allow / yolo modes, allowlist patterns) and the verify-it-works step.

Triggers

The skill’s description lists the verbatim phrasings the agent will match. The most common:

• “help me set up core-agent”
• “configure core-agent for my repo”
• “what should my AGENTS.md look like”
• “what permissions should I use”
• “how do I add a skill”

Any phrasing implying first-time configuration or extending an existing setup.

References

The skill body is short (a runbook). It fetches deeper content on demand from three reference files:

• agents-md-patterns.md — AGENTS.md authoring: the three-layer instruction stack, minimum viable shape, patterns that work, anti-patterns, model-specific quirks (Pro vs Flash), iteration approach.
• skills-and-mcp.md — extending the tool surface: when to write a skill vs an AGENTS.md rule, SKILL.md schema, description-as-trigger pattern, MCP server wiring + namespacing.
• permission-posture.md — the three modes, pattern grammar, in-session “always allow” persistence, bash denylist, path-scope safety floor, common configurations.

When to invoke vs read the docs

The skill IS the docs in workflow form. Pick the surface that matches your read intent.

Installing

See Skills library → Installing.

Quick version:

cp -r /path/to/core-agent/SKILLS/cli-setup .agents/skills/

Verify with /skills in the TUI.

Adapting

The skill teaches core-agent’s defaults. If your organization has a stricter standard (e.g., “every project must use anthropic-vertex with mode=ask”), fork the bundle:

cp -r .agents/skills/cli-setup .agents/skills/cli-setup-acme
# Edit SKILL.md to encode your org's defaults

The agent will now invoke cli-setup-acme instead of the generic version (assuming the description matches the same triggers — keep the trigger phrasings or extend them).

Where to go next

• Interactive quickstart — the static-docs version of the same walkthrough
• System instructions — the prescriptive content the skill’s agents-md-patterns.md reference is built from
• Reference → Skills — SKILL.md schema, discovery, allow/deny lists