compound

To capture team learnings from a bug fix, you can generate frontmatter-tagged markdown files within your repository's docs/solutions/ directory. This records the problem, root cause, and prevention steps for future discoverability without re-reading long conversations.

Writing a post-mortem doc that agents can discover involves creating frontmatter-tagged markdown files in docs/solutions/ and adding pointers to AGENTS.md or CLAUDE.md. This re-injects codified lessons into future planning and agent-assisted review runs.

Yes, you can write solution docs directly from your current conversation using a lightweight capture mode. This allows you to quickly generate and update markdown files without running parallel research passes or overlap scoring.

Yes, generating solution docs in full mode performs overlap scoring against existing files in the docs/solutions/ directory. This ensures new learnings are categorized correctly and prevents duplicating previously captured knowledge.

You structure repository-scoped learnings by applying a provided frontmatter schema to markdown files, selecting a category and slug, and writing them under docs/solutions/. This formalizes decisions and fixes into durable, structured knowledge.

No, you do not need external models to run discoverability checks on solution docs. The process checks or bootstraps the docs/solutions/README.md and writes markdown files locally without invoking external models.