agentsclimarketplace

Skill creator

Skill Anoxxx/skillcache/examples/skill-creator

MRU attention protocol for skills that learn from every use — zero compute

Install
npx -y skills add Anoxxx/skillcache --skill skill-creator

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

2 things to look at

  • no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
  • 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

Create new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit or optimize an existing skill, run evals to test a skill, benchmark skill performance, or optimize a skill's description for better triggering accuracy.

The file declares its own license as Original by Anthropic (Apache 2.0). Adapted here as an adaptive skills example.. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.

SKILL.md

3.6 KB, as published. Nobody here has run it

Skill Creator

A skill for creating new skills and iteratively improving them.

Process Overview

  1. Decide what the skill should do and roughly how
  2. Write a draft of the skill (SKILL.md + any scripts/resources)
  3. Create test prompts and run the agent with the skill on them
  4. Evaluate results qualitatively (does it feel right?) and quantitatively (evals)
  5. Rewrite based on feedback
  6. Repeat until satisfied
  7. Expand test set and try at larger scale

Your job is to figure out where the user is in this process and help them progress.

Capture Intent

Start by understanding:

  1. What should this skill enable the agent to do?
  2. When should this skill trigger? (what user phrases/contexts)
  3. What's the expected output format?
  4. Should we set up test cases to verify?

Skills with objectively verifiable outputs (file transforms, data extraction, code generation) benefit from test cases. Skills with subjective outputs (writing style, art) often don't need them. Suggest the appropriate default based on skill type, but let the user decide.

Write the SKILL.md

Required frontmatter fields:

  • name: Skill identifier (lowercase, hyphens for spaces)
  • description: When to trigger + what it does. This is the primary triggering mechanism.

The description should be slightly "pushy" — include explicit trigger phrases AND negative cases ("NOT for: ..."). Agents have a tendency to undertrigger skills, so descriptions need to actively encourage activation.

Body Structure

Good skills follow this pattern:

  1. One-line summary of what the skill does
  2. Decision tree or workflow — the core logic
  3. Examples — concrete input/output pairs
  4. Guidelines — constraints, edge cases, anti-patterns
  5. Reference files — pointer to scripts/ and resources/ if they exist

Keep the body concise. If detailed reference material is needed, put it in reference/ files and point to them from the body. The body itself should be loadable without excessive token cost.

Test and Iterate

For each iteration:

  1. Run 5-10 test prompts through the skill
  2. Review outputs — does the skill trigger correctly? Are outputs useful?
  3. Identify failure modes — undertriggering, wrong outputs, missing edge cases
  4. Update SKILL.md and re-test

Description Optimization

The skill's description field is the primary trigger mechanism. Optimize it by:

  • Including common user phrasings that should activate the skill
  • Including "NOT for: ..." to prevent false triggers
  • Testing trigger accuracy across diverse prompts
<!-- adaptive-skills v0.1 -->

Adaptive Heuristics

This skill maintains a HEURISTICS.md alongside this file. Load it at skill activation. After completing the task:

  1. Did anything unexpected happen? (failure, workaround, surprise, user correction) → Append a one-line lesson to the TOP of HEURISTICS.md.
  2. Did you rely on an existing heuristic? → Move it to the TOP (MRU promotion).
  3. Does a heuristic contradict this SKILL.md? → SKILL.md wins. Note the conflict in the heuristic entry.

Keep heuristics ≤ 20 entries. When full, the bottom entry is the eviction candidate.

<!-- /adaptive-skills -->

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.