Stats
Actions
Tags
From pro-workflow
Trains a skill's SKILL.md by running a budget-optimized offline loop over accumulated learn-rule corrections, validating patches against held-out trajectories before overwriting the file.
How this command is triggered — by the user, by Claude, or both
Slash command
/pro-workflow:skill-optimizeThe summary Claude sees in its command listing — used to decide when to auto-load this command
# /skill-optimize - SkillOpt-flavored offline training Run an offline, budget-capped optimization loop over a skill's accumulated `learn-rule` trajectories. Proposes bounded patches via an optimizer LLM, validates each candidate against a held-out portion of the same trajectories, and overwrites SKILL.md only when the candidate strictly improves the weighted score. ## Quick Start ## What it does 1. Pulls recent `learnings` rows scoped to the skill slug (or global) 2. Splits them into train + validation (~25% holdout, freezes validation set) 3. Runs `epochs` x `minibatches` rounds of: ...
Run an offline, budget-capped optimization loop over a skill's accumulated learn-rule trajectories. Proposes bounded patches via an optimizer LLM, validates each candidate against a held-out portion of the same trajectories, and overwrites SKILL.md only when the candidate strictly improves the weighted score.
/skill-optimize <slug> [--epochs 3] [--budget-usd 0.50]
learnings rows scoped to the skill slug (or global)epochs x minibatches rounds of: reflect → aggregate → clip → apply → evaluate → gate~/.pro-workflow/STOP), no improvement, or epochs donelearnings rows for the slugANTHROPIC_API_KEY (or OPENAI_API_KEY / OPENROUTER_API_KEY / FIREWORKS_API_KEY with matching --optimizer-provider)npm run build has been run in the pro-workflow plugin directory at least once/skill-optimize pro-workflow
/skill-optimize wiki-research-loop --budget-usd 1.0 --epochs 5
/skill-optimize wrap-up --optimizer-model claude-opus-4-7 --evaluator-model gpt-4o-mini
The third example mixes providers. The CLI infers the provider from the model id (claude-* → anthropic, gpt-* / o* → openai), so you do not need --evaluator-provider openai for gpt-4o-mini. Pass an explicit --optimizer-provider / --evaluator-provider to override inference.
See skills/skill-optimizer/SKILL.md for full mechanics, defaults, and the SkillOpt provenance.
npx claudepluginhub rohitg00/pro-workflow --plugin pro-workflow/skill-optimizeOptimizes a skill's description for trigger accuracy by generating eval queries, running a train/test split optimization loop via Python, and applying the best version after user confirmation.
/improveIteratively improves a skill folder based on feedback, running baseline review, diagnosing issues like triggers or output quality, and applying targeted fixes.
/eval-evolveEvolves SKILL.md via multi-round Ralph loop chaining /eval-skill → /eval-improve until target score or max iterations. Installs ralph-wiggum plugin if needed.
/improve-skillsAnalyzes LEARNINGS.md for skill improvement opportunities from execution logs, failure rates, and user evaluations; prioritizes by frequency-impact-ease score; proposes and optionally implements fixes with validation.
/evolveEvolves skill .md files by integrating validated lessons from usage data, generating diffs for user approval. Also supports --rollback, --export, and --import.
/refactor-skillsAnalyzes and refactors skills by consolidating duplicates, optimizing, and removing redundancy.