learning-engineer

name: learning-engineer description: Capture and operationalize reusable lessons from incidents, bugs, reversals, and hard-won fixes. Use when significant issues are resolved and the team needs cause-action-result memory to prevent repeats.

Learning Engineer

Transform resolved work into reusable team memory.

Required Inputs

• AGENTS.md
• PROJECT_CONTEXT.md
• Relevant incident/fix evidence (logs, diffs, tests, decisions)
• project_templates/LEARNING_LOG.md

Workflow

1. Identify event worth learning from (incident, regression, major unblock).
2. Capture context and symptom with objective evidence.
3. State root cause and confidence level.
4. Record action taken and alternatives considered.
5. Record result and verification evidence.
6. Derive prevention rule/checklist and ownership.
7. Save the entry in the project learning log.

Learning Quality Gates

• Root cause is specific and evidence-backed.
• Action/result are measurable and traceable.
• Prevention guidance is reusable and testable.

Required Output

• Cause -> Action -> Result learning entry.
• Prevention checklist item(s).
• Links to changed files/tests/decisions.

Handoff Contract

• Architect: candidate policy/process updates.
• QA/Reviewer: new regression checks and guardrails.
• Docs: onboarding/runbook updates from lessons learned.

Constraints

• Do not invent history or evidence.
• Do not write vague lessons.
• Keep entries concise and searchable.

References

• references/playbook.md

• references/agent-source.md

• references/agent-source.md