skill-reviewer

You can audit AI agent skills for local-model fitness by analyzing their metadata, prose, and structure against best practices for portability, determinism, and context budget limits. This process identifies risks like vendor overfit and large model assumptions.

Best practices for preventing vendor lock-in include ensuring portability, implementing progressive disclosure, and maintaining determinism. Reviewing skills against these standards detects excessive vendor-specific tuning and large model assumptions before integration.

To check if an AI skill will exceed your context budget, analyze its structure and prose to identify context budget overruns. The review process assesses metadata and content to flag risks related to excessive context usage.

Yes, you can evaluate determinism in AI agents by analyzing skill metadata, prose, and structure against vendor-aware best practices. The review identifies risks related to vendor-specific tuning and provides structured findings and scores for assessment.

You should run a skill audit before integrating new components into your AI agent to ensure they meet project standards for performance and compatibility. This prevents vendor lock-in and verifies suitability for local or smaller models.