Skills

A collection of agent skills for coding assistants, following the Agent Skills specification. Works with Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI, Amp, OpenCode, and any other agent that reads the SKILL.md format.

Install

Install all skills:

npx skills add lukebennett88/skills

Install a single skill:

npx skills add lukebennett88/skills/<skill-name>

The skills CLI symlinks skills into each agent's directory, so npx skills update keeps them current.

Skills

Skill	Description
lb-code-review	Two-axis review of a diff — standards compliance and spec fidelity — via parallel sub-agents
lb-code-style	JS/TS and React readability defaults — early returns, lookups over switch, Map/Set for lookups, plain loops over clever reduce
lb-deslop	Make prose specific, plain, and readable — remove AI tells, filler, and formulaic structure
lb-domain-modeling	Pin down project vocabulary and record durable decisions in AGENTS.md and docs/ as the model sharpens
lb-grill-with-docs	One relentless interview to sharpen a plan, capturing terms and decisions into project docs as they crystallise
lb-grilling	Relentless one-question-at-a-time interview to stress-test a plan or design
lb-implement	Build a spec or ticket, using TDD at agreed seams, then hand off to code review
lb-prototype	Throwaway UI or logic prototype to answer a design question fast
lb-tdd	Red-green-refactor discipline — seams, good vs bad tests, and mocking rules
lb-to-spec	Synthesize the current conversation into a spec (PRD) published to the issue tracker
lb-to-tickets	Break a spec or plan into tracer-bullet tickets with blocking edges on the issue tracker
lb-wayfinder	Chart a big chunk of work as a map of investigation tickets, resolved one session at a time

Planning pipeline

The engineering skills form a pipeline, each step handing off to the next:

1. lb-grill-with-docs (or lb-wayfinder for work too big for one session) — sharpen the plan
2. lb-to-spec — synthesize it into a spec
3. lb-to-tickets — break the spec into tickets
4. lb-implement — build a ticket
5. lb-code-review — review the result

lb-grilling, lb-domain-modeling, lb-prototype, and lb-tdd are supporting skills the pipeline invokes along the way.

These ten skills are forked from mattpocock/skills (MIT).

Structure

skills/
  <skill-name>/
    SKILL.md          # required — frontmatter (name, description) + body
    references/       # optional — heavy docs, loaded on demand
    scripts/          # optional — executable tools
template/
  SKILL.md.template   # starter template for new skills
scripts/
  validate.mjs        # spec validator (runs in CI)
  validate.test.mjs   # validator tests (run in CI)

Writing a new skill

1. Copy template/SKILL.md.template to skills/lb-<skill-name>/SKILL.md. All skills are prefixed lb- so /lb surfaces them and names never collide with skills from other sources.
2. Set name (must match the directory) and description (triggering conditions only, start with "Use when").
3. Keep the body under 500 lines; push heavy material into references/ or scripts/.
4. Test it: run the scenario without the skill, confirm the failure, then confirm the skill fixes it.
5. Validate:

node scripts/validate.mjs
node --test scripts/validate.test.mjs

See AGENTS.md for full conventions.

License

MIT