skill-optimizer

Fix unreliable Agent Skill behavior by applying evidence-based diagnosis to real-world conversation and execution traces, identifying root causes, and generating reviewable patches that update triggers, inputs, and workflow steps to prevent repeat failures.

Evidence-based skill hardening converts observed failures and corrections into safer, clearer, reviewable updates. You need it when an existing Agent Skill requires debugging, generalization, or durable changes derived from execution evidence.

Package an Agent Skill by organizing skill folders, associated references, scripts, and documentation into a distributable structure, ensuring the hardened skill and its validation cases are ready for deployment.

Benchmark and validate an Agent Skill by producing validation cases and trigger evaluation suggestions, verifying that evidence-based patches improve behavior and catch regressions in the updated skill workflow.

Yes, you can generate reviewable patches for SKILL.md sections while preserving safety constraints and appropriate authority gates. Proposed edits are small, auditable, and include updated inputs, triggers, workflow steps, and outputs.