Agent-Skills.md

Agent Skills: optimize-description

Run automated description optimization loop with train/test split. Use when a skill's trigger accuracy is low or the description needs refinement.

UncategorizedID: bluewaves-creations/bluewaves-skills/optimize-description

Repository

bluewaves-creations/bluewaves-skills
bluewaves-creationsLicense: MIT
1

Install this agent skill to your local

pnpm dlx add-skill https://github.com/bluewaves-creations/bluewaves-skills/tree/HEAD/plugins/skills-factory/skills/optimize-description

Skill Files

Browse the full folder contents for optimize-description.

Download Skill

Loading file tree…

plugins/skills-factory/skills/optimize-description/SKILL.md

Skill Metadata

Name
optimize-description
Description
Run automated description optimization loop with train/test split. Use when a skill's trigger accuracy is low or the description needs refinement.

Optimize a skill's description through iterative eval-and-improve cycles.

$ARGUMENTS

If $ARGUMENTS is provided, use it as the skill path. Otherwise, ask the user which skill to optimize.

Steps

  1. Locate the skill directory. Resolve $ARGUMENTS to an absolute path if provided.

  2. Check for eval queries in .skill-eval/evals/ -- if missing, scaffold them first.

  3. Ask the user to review trigger queries in eval_review.html (generated by the skill-eval pipeline).

  4. Run the optimization loop:

    python3 ${SKILL_ROOT}/../skill-shaper/scripts/run_loop.py \
  --eval-set <queries> \
  --skill-path <dir> \
  --max-iterations 5 \
  --holdout 0.4

  5. Open the live HTML report (generate_report.py) showing iteration progress.

  6. When complete, display the best description and its test score.

  7. Ask the user to confirm before applying the new description to SKILL.md.