retrospective-codify

To convert post-task learnings into reusable lint rules, you summarize the initial failure, final solution, and bridging insight from a completed task. This structured prompt generates deterministic ast-grep rules, lint checks, or AGENTS.md entries to prevent repeating mistakes.

The best way to create language-agnostic ast-grep rules from past mistakes is to extract the bridging insight between the initial failure and final solution. This insight is then codified into a new or updated skill artifact.

Yes, you can use retrospective learnings to generate engineering guidelines. By structuring task completion insights into deterministic rules, the skill creates or augments AGENTS.md entries and lint-style checks applicable across projects or teams.

To codify task insights into ast-grep rules, you must provide structured prompts detailing the initial failure, the final solution, and the bridging insight. This specific context is required to output a valid skill artifact.

Yes, codifying retrospective insights works across different programming languages. The skill outputs language-agnostic rules and guidelines, ensuring that learnings from one task can be applied as ast-grep rules or AGENTS.md entries across diverse environments.

You should structure prompts with failures and solutions for retrospective codification because the skill requires this specific context to accurately extract the bridging insight. Without this structured input, it cannot generate deterministic, reusable rules.