lucidpaths/claude.md-workflow/pick-skills

Claude Code and agent skill by lucidpaths/claude.md-workflow

Trust tier 2
  • community
LLM review pending
Source check pending

Checking that the SKILL.md file is still in the repository.

Choose the skills a task needs before starting it — read every skill this session can invoke, match them to the task's phases, state the chain and why, then run it, or say plainly that none fits. Use when the user runs /pick-skills (optionally with a task), says "which skill fits", "use the right skills for this", or starts a broad task without naming a method. Not needed when the user already named the skill.

Install

npx vskill@latest install lucidpaths/claude.md-workflow/pick-skills

SKILL.md

The same SKILL.md was published earlier as lucidpaths/claude.md-workflow/pick-skills.

Pick skills — the right discipline for the task, chosen on purpose

Skills only help when the right one fires. Automatic loading is a guess made from one sentence of description; this skill makes the choice explicit: read what is available, judge it against the task, say what will run and why, then run it.

1. Pin the task

  • One sentence: what must exist or be true at the end. Without a task in the invocation, take the open request of the

conversation. ASK once when there is no open request, or it yields two different end-state sentences: "What must exist or be true at the end: <A> or <B>?" → an end state named: pin it / nobody to ask: report the readings to whoever dispatched you and stop before choosing skills.

  • Its phases: understand, plan, build, verify, review, ship, record. Most tasks touch two or three.
  • Its stakes: reversible or not, data or money at risk, shared or public output, size.

2. Read the bench

  • List every skill this session can invoke: the skill list the harness shows, plus the skill folders of the project

and the user (.claude/skills/*/SKILL.md) and of installed plugins. Names and descriptions first.

  • Open a SKILL.md in full only when its description leaves the fit undecided: two candidates for one job, or a

description that does not say when to use it.

  • Only what this session lists can be invoked. A skill remembered from elsewhere does not exist here until it is listed.

3. Choose

For each phase, at most one skill: the most specific one that covers it.

  • Specific beats general: a skill whose description names the task's system, artifact or kind of work beats one that

names only a phase or a method; a project's skill beats a personal one for the same job.

  • A skill is not used when its own description excludes the task, or when it requires artifacts or agents that the

pinned end state does not need (a team planning procedure for a one-line fix). Say so.

  • Skills compose in phase order: plan, then build, then verify, then record. Two skills that still claim the same job in

the same phase after the tests above are a conflict. ASK on a conflict: "For <phase>, run /<A> (<what it does>) or /<B> (<what it does>)?" → A: chain A, list B under NOT USED / B: chain B, list A under NOT USED / nobody to ask: report both with their descriptions to whoever dispatched you, and run only the phases that do not depend on the choice.

  • A phase no skill covers is a gap. List it under GAPS and do that phase without a skill, under the rules the session

already carries (the user's and the project's instructions); never stretch a near-miss skill over it.

  • A skill that writes to shared systems (trackers, repositories, messages, bookings) runs only within what the user has

authorised.

4. Say it, then run it

State the chain with the closing block below, then invoke the first skill and the next ones as the work reaches their phase. If the user asked only which skill fits, stop after the block.

Closing block (required)

TASK <one sentence>
CHAIN 1. /<skill> <phase> <what it contributes to this task>
2. ...
NOT USED /<skill> <why it looked relevant, and why it does not apply>
GAPS <phases no skill covers, or "none">

Security

  • Security scanPassed, 100 of 100Pattern checks on the skill's files.
  • Model reviewNot runA model review is scheduled for the next rescan.

Quality evaluation

  • Task evaluationNot runNo evaluation yet. Run one in Skill Studio.

Source

Popularity

Repository stars
8
7-day trend score
0

Stars and forks are for the whole GitHub repository, not this skill alone. Not refreshed yet.

Works with

Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.

  • Claude Code
  • Cursor
  • GitHub Copilot
  • Windsurf
  • Codex
  • Gemini CLI
  • Cline
  • Amp
  • Roo Code
  • Goose
  • Continue
  • Kilo Code
  • Junie
  • Trae
  • OpenHands
  • Qwen Code
  • AI Agent
  • Tabnine Chat
  • Sourcegraph Cody
  • Amazon Q
  • Firebase Genkit
  • Sweep
  • Devin
  • Phind
  • Mentat
  • Claude Desktop
  • Aider
  • GPT Pilot
  • Plandex
  • bolt.new
  • v0
  • Replit Agent
  • Anse
  • TypingMind
  • BetterChatGPT
  • LibreChat
  • Warp AI
  • Pieces
  • Fig AI
Embed badge
vskill VERIFIED badge
[![vskill VERIFIED](https://verified-skill.com/api/v1/skills/lucidpaths/claude.md-workflow/pick-skills/badge)](https://verified-skill.com/skills/lucidpaths/claude.md-workflow/pick-skills)

Something wrong with this skill or its page? Tell us.