deepwork_jobs.learn

Analyze conversation history from completed job runs to identify patterns and inefficiencies, classify them as generalizable or bespoke, and update instruction files automatically to improve instructions for future executions.

An AGENTS.md file captures project-specific knowledge extracted from job runs. Generate it in the deepest common folder after a job completes to package reusable guidance and improve future DeepWork executions.

Run the Learn skill to analyze the completed job's conversation history, extract generalizable improvements, and update instruction files and AGENTS.md in the deepest common folder, then prompt a sync to finalize modifications.

Yes, analyzing conversation history distinguishes generalizable instruction improvements from bespoke run-specific insights, ensuring only reusable guidance updates instruction files while unique observations are surfaced separately.

Yes, after instruction files and AGENTS.md are updated with extracted learnings, the system prompts a manual sync to ensure all modifications are ready and properly applied for future runs.

Automate extraction by analyzing conversation histories to identify successful patterns and confusions, classifying learnings, and updating AGENTS.md and instruction files to package reusable guidance for future work.