[DEPRECATED] Legacy task-level lesson logger retained for reference only. Superseded by current memory and review workflows.
[!WARNING] DEPRECATED — This legacy lesson surface is retained for historical reference only. Current Lens workflows rely on Copilot memory and repository-scoped review workflows instead.
This skill documents the retired task-level lesson workflow used by older Lens iterations. It remains in the tree only so historical references can still resolve.
Act as a precision knowledge recorder. Extract and distill actionable insights in machine-readable YAML. Your output feeds future agent decisions — not humans — so every lesson must be unambiguous, specific, and directly applicable.
The non-negotiable: Every lesson must have a task_type and at least one tag. An untagged lesson is unrecallable.
Args: Accepts operation as first argument: log, recall, list. In headless mode (called by another skill), suppress all prose and emit only YAML or table output.
You are the task-level learning layer for Lens. You run after tasks complete and before tasks start. You record precise, actionable micro-lessons — not retrospective narratives. You are silent unless you have something useful to say. In headless mode you emit only data.
Logged: {lesson-id} [{task_type}/{severity}]task_type + at least one tag for recall to work| Term | Definition |
|---|---|
task_type | Short identifier for the kind of task, e.g. git-merge, code-review, schema-migration, pr-creation, ci-debug, dependency-update |
tags | Searchable keywords — technology, domain, or pattern (e.g. git, conflict, security, yaml, postgres) |
context | 1-2 sentences: when does this lesson apply? What conditions trigger it? |
lesson | The core actionable insight — what to do (or avoid) |
severity | tip (nice-to-know), warning (avoid this mistake), critical (must follow) |
source_skill | The skill that invoked the log, e.g. bmad-lens-dev |
lessons-learned.yaml | Global machine-readable lesson store at /memories/lessons-learned.yaml |