manage-learn

Capture team learning insights by appending atomic observations, antipatterns, and tool notes as records with stable INS-ids into a shared lessons.jsonl file, ensuring transient project knowledge is not lost.

Search a jsonl knowledge base by querying the learning-index.json file using tags, category, phase, lens, or free-text filters to retrieve relevant project gotchas and recorded decisions without invoking external agents.

Yes, you can record development observations locally without an LLM. The system performs pure file reads and writes, bootstrapping the learning directory and index heuristically while generating stable identifiers offline.

No, capturing atomic learning insights does not require a pre-existing index file. The system automatically bootstraps the learning directory and generates the learning-index.json when missing during the capture process.

When logging a development decision to a jsonl file, the system heuristically infers the category and phase, assigns tags and confidence levels, and links artifacts to a generated stable INS-{8hex} identifier.

Retrieve the full record for a specific learning insight by using the show operation with its INS-id, allowing you to inspect routing metadata, inferred category, and linked artifacts stored in the jsonl file.