Skills

A repository of custom agent skills following the Agent Skills open format. Each skill packages specialised procedural knowledge — structured instructions, templates, and domain context — that any compatible LLM agent can load on demand.

Skills are LLM-agnostic and agent-agnostic. The same skill works across Claude, GPT, Gemini, and any other agent that supports the Agent Skills format (Cursor, Claude Code, Codex, and others).

Install (for users)

Install this package to copy all skills into your local agent skills directory:

npm install @daviesayo/rocket-skills

Or install without adding it to a project:

npx @daviesayo/rocket-skills@latest

What gets installed where

Each skill is copied into ~/.agents/skills/ as its own folder. For example, after installing you get:

• ~/.agents/skills/running-consultant-engagements/ — the skill folder and its contents (e.g. SKILL.md, templates/)

Agents that read skills from ~/.agents/skills/ (e.g. Cursor with the right config) will then see and use these skills automatically.

Uninstall

Removing the npm package does not remove the copied files from ~/.agents/skills/. To remove a skill, delete its folder:

rm -rf ~/.agents/skills/running-consultant-engagements

(Repeat for any other skill names you want to remove.)

What is a skill?

A skill is a folder containing a SKILL.md file. The file has two required pieces: a short YAML frontmatter block (name + description) and a Markdown body with the actual instructions. When an agent detects that a user request matches a skill's description, it reads the full SKILL.md into context and follows the workflow inside.

Skills can also bundle supporting files — templates, reference docs, scripts — which the agent loads on demand as needed. This keeps context lean at startup while making richer resources available during execution.

See the Agent Skills specification for the full format reference.

How the skills in this repo were built

Skills here were authored using the TDD-for-documentation methodology from Superpowers, specifically the writing-skills skill.

The process treats skill creation like test-driven development:

1. RED — run a baseline scenario with a subagent and document exactly how it fails or goes wrong without the skill
2. GREEN — write the SKILL.md addressing those specific failure modes; verify the agent now complies
3. REFACTOR — find remaining loopholes in the agent's reasoning, close them, re-verify

Key authoring principles from that methodology, applied here:

• Description = triggering conditions only. The description field tells the agent when to load the skill, not what the skill does. Summarising the workflow in the description creates a shortcut the agent takes instead of reading the full instructions.
• One artifact at a time. Workflows with sequential dependencies enforce explicit review gates — the agent produces one output, waits for approval, then continues.
• Red flags over vague guidance. Each skill includes a concrete list of failure modes and what the correct behaviour is, so the agent has something specific to check against.
• Progressive disclosure. Heavy reference material lives in separate files under skill/templates/ or skill/references/, loaded only when needed rather than bloating the main SKILL.md.

How this repo is organised

skills/
└── <skill-name>/
    ├── SKILL.md       ← required; frontmatter + instructions
    └── ...            ← supporting files (templates/, references/, etc.)

The skills/<skill-name>/ folder is the unit of deployment. When users install via npm, a postinstall script copies each such folder into ~/.agents/skills/<skill-name>/ on their machine.

Available skills

| Skill | Description | |-------|-------------| | running-consultant-engagements | Orchestrates a structured consulting engagement workflow — from raw discovery inputs to 7 cross-referenced artifacts, generated one at a time with review gates between each step |

Maintainer

Adding a new skill

1. Create skills/<your-skill-name>/ and add a SKILL.md at the root of that folder (see the spec for the required frontmatter format).
2. Add any supporting files the skill references (e.g. templates/, references/) inside skills/<your-skill-name>/.
3. Add a row to the “Available skills” table above.
4. (Optional) Validate with the skills-ref tool: skills-ref validate ./skills/<your-skill-name>/.

No changes to package.json are needed. The postinstall script copies every directory under skills/ that contains a SKILL.md.

Publishing to npm

One-time: Create an account at npmjs.com if needed, then from the repo run:

npm login

Each release:

1. Bump the version (e.g. npm version patch).
2. Publish:

   npm publish

3. Push and push tags:

   git push && git push --tags

Users get updates with npm install @daviesayo/rocket-skills or npm update @daviesayo/rocket-skills. This package is scoped; publish with npm publish --access public.