agentsclimarketplace

Write a skill

Skill slogsdon/skills-meta-utils/skills/write-a-skill

Meta-utility skills for Claude Code: compress prompts to save tokens, delegate work to local Qwen via LM Studio.

Install
npx -y skills add slogsdon/skills-meta-utils --skill write-a-skill

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Guide writing a new skill file following Shane's authoring conventions — principle-first, anti-patterns before steps, _lib/ references over inline duplication. Use when invoked as /write-a-skill or when Shane says "create a new skill for X", "add a skill", "write a skill that does X".

SKILL.md

2.5 KB, as published. Nobody here has run it

Skill: /write-a-skill [name]

A skill is a short, opinionated document — write the minimum that removes ambiguity, then stop.

Don't: embed blocks that appear in other skills — extract them to _lib/. Don't lead with numbered steps before stating what the skill is for.

Conventions

Length targets:

  • Simple delegation skill (Qwen, one task): under 40 lines
  • Complex ritual skill (multi-step, fallback): under 80 lines
  • If over limit, find what to extract to _lib/

Structure (in order):

  1. Frontmatter: name, description (include trigger phrases and "Do NOT use for" guards)
  2. H1 with skill name
  3. Principle line — one sentence stating what the skill produces and why it's valuable; not "this skill does X" — just the thing itself
  4. Anti-patterns — 1–3 **Don't:** lines naming specific failure modes and the right alternative
  5. Steps or happy path
  6. Fallback (if tool-dependent) — kept short via _lib/ references

Principle line pattern: Combine WHAT and WHY into one sentence. Examples:

  • /connect: "Surface the non-obvious bridges between two concepts the vault hasn't linked yet — the value is in connections that surprise, not the ones you'd expect."
  • /trace: "Map the arc of how Shane's thinking on a topic evolved — not the current position, but the path that led there and the inflection points along the way."

_lib/ candidates:

  • Any block >10 lines that appears in more than one skill
  • Tool invocation protocols (Qwen delegation loop, obsidian commit sequence)
  • Save/commit sequences

Process

  1. Determine the skill name and its trigger phrases.
  2. Write the description frontmatter: trigger phrases + "Do NOT use for..." guard.
  3. Write the principle line — one sentence, WHAT + WHY combined.
  4. Write 1–3 anti-patterns using the **Don't:** format; name the right alternative.
  5. Write steps — collapse any step that just restates what to do without showing how. If the skill depends on an external tool (Qwen, MCP), add a Fallback section.
  6. Identify _lib/ candidates — any block you've seen in another skill.
  7. Count lines. If over target, extract one more thing or delete it.
  8. Save to plugins/shane/<plugin>/skills/<name>/SKILL.md.

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.