meta-discovery

Pattern detection for reusable skills works by applying a two-phase evaluation: first analyzing if a multi-step pattern exists and is likely to recur, then deciding whether to propose creating a new skill to capture that workflow.

You should evaluate task workflows for skill creation immediately after finishing a substantive interaction. This is when the context of multi-step patterns is fresh, allowing the evaluation mechanism to accurately assess if the workflow is likely to recur.

Yes, the suggestion phase automatically proposes new skills by deciding if a pattern should be captured and generating a structured recommendation using a {prefix}-{name} format. This ensures consistent skill naming and structured guidance for project-wide skill creation.

No dependencies are required to track conversation and skill identifiers. The skill evaluation process utilizes provided scripts to track identifiers directly within the workflow, operating independently without external components or libraries.

The best way to evaluate if a coding pattern should become a reusable skill is to apply a two-phase evaluation: first check if a multi-step pattern exists and is likely to recur, then decide whether to propose capturing it as a new skill.