lesson-capture

To capture lessons learned, this Skill extracts actionable insights from completed tasks and appends them to AGENTS.md. It acts as a centralized knowledge base by validating evidence and requiring user approval before publishing to the lessons section.

Documenting project decisions and failures is handled by classifying validated lessons as PATTERN, DECISION, GOTCHA, or FAILURE. The Skill tags these insights with a suitable scope and appends them to AGENTS.md after batch approval.

Gate-based eligibility for logging lessons works by enforcing evidence, user approval, and scope validation before appending. It evaluates task candidates to ensure only quality, actionable insights are saved to AGENTS.md.

Yes, you can use this to extract knowledge from conversations and sessions. The Skill applies its capture flow to tasks, conversations, and sessions within a repository to validate evidence and append approved lessons to AGENTS.md.

Yes, user approval is required before lessons update AGENTS.md. The Skill prepares a batch of validated lessons and enforces evidence and scope validation, but you must approve them before they are appended to the knowledge base.

Using AGENTS.md for project knowledge management maintains a centralized, validated repository within your codebase. Unlike generic notes, it enforces evidence and scope validation through quality gates to ensure only actionable lessons are stored.