From skill-improver
Iteratively reviews and fixes a Claude Code skill until it meets quality standards. Accepts a skill name or path and optional --max-iterations flag.
How this command is triggered — by the user, by Claude, or both
Slash command
/skill-improver:skill-improver <SKILL_NAME_OR_PATH> [--max-iterations N]This command is limited to the following tools:
The summary Claude sees in its command listing — used to decide when to auto-load this command
# Skill Improver ## Step 1: Resolve Skill Path The user provided: `$ARGUMENTS` **Resolution rules:** 1. If input ends with `/SKILL.md` and file exists, use directly 2. If input is a directory containing `SKILL.md`, use that path 3. Otherwise, search for matching skills: Filter results to find skills matching the user's input (by skill name or path substring). - **Multiple matches:** Ask the user to choose - **No matches:** Report available skills - **Single match:** Proceed with that path Store the resolved absolute path to the skill directory in `RESOLVED_SKILL_PATH`. ## Step 2: I...
The user provided: $ARGUMENTS
Resolution rules:
/SKILL.md and file exists, use directlySKILL.md, use that pathGlob(pattern="**/SKILL.md")
Filter results to find skills matching the user's input (by skill name or path substring).
Store the resolved absolute path to the skill directory in RESOLVED_SKILL_PATH.
Run the setup script with the resolved path:
"${CLAUDE_PLUGIN_ROOT}/scripts/setup-skill-improver.sh" "<RESOLVED_SKILL_PATH>" [--max-iterations N]
Substitute <RESOLVED_SKILL_PATH> with the actual path. Add --max-iterations N only if specified.
Apply the skill-improver:skill-improver skill to iteratively improve the target skill. The loop continues automatically via the stop hook until completion.
npx claudepluginhub trailofbits/skills --plugin skill-improver/improveIteratively improves an existing skill based on real-world feedback, test results, or edge cases. Diagnoses root causes and applies targeted fixes.
/autoRuns an autonomous improvement loop on a skill — scores it, applies deterministic patches, re-scores, and keeps or reverts changes until plateau or target is reached.
/eval-evolveRuns an autonomous multi-round evolution loop on a SKILL.md file, chaining evaluation and improvement until PASS verdict or iteration limit. Requires the ralph-wiggum plugin.
/corgi-improve-skillRefines or creates an agent skill from conversation evidence. Accepts a skill path and gap description; edits existing skills, scaffolds new ones, or summarizes gaps without changes.
/skill-evalTests a skill by running parallel evaluation cases with and without the skill, compares outputs against assertions, and iterates to improve skill quality based on results.
/auto-improveAutonomously improves SKILL.md files using a quality-scoring loop: measure, research, improve, keep or revert. Also supports baseline scoring, session reporting, and audit-findings improvement.