update-skills

To capture repository learnings, you record non-obvious constraints, recurring bugs, and architectural rules as concise, actionable guidance with concrete examples. This ensures important repository knowledge persists beyond a single debugging session for future reuse.

Document debugging pitfalls as reusable instructions when a significant pattern, recurring failure mode, or correction is discovered. This preserves valuable architectural constraints and domain knowledge that should persist beyond the current investigation.

Choosing the right destination file requires deciding whether the insight belongs in an existing learning, a new instruction, or a standalone skill file. This ensures repository knowledge is placed in a durable, searchable form for future invocation.

The best way to write skill files for recurring architectural constraints is to create or update files with clear guidance, concrete examples, and a description that makes future invocation obvious. This captures reusable learnings effectively.

Writing repository knowledge into instructions works for non-obvious pitfalls by preserving the discovered correction in a durable format. This prevents future sessions from repeating the same investigation when encountering similar failure modes.