conversation-logger

Conversation Logger

Write a conversation log for every task you complete. This is non-negotiable.

When

• Before returning to the orchestrator. The log is part of your task deliverable.
• One log file per task. Not per session, not per feature — per task.

Where

.github/conversations/{YYYY-MM-DD}/{seq}-{role}-{task-slug}.md

• {YYYY-MM-DD} — today's date
• {seq} — next available sequence number in that date directory (read existing files to determine)
• {role} — your role in lowercase (engineer, qe, pm, debugger)
• {task-slug} — short kebab-case task description (e.g., implement-auth, run-tests, trace-timeout)

Format

See .github/conversations/SCHEMA.md for the full specification. Key points:

1. YAML frontmatter — task, agent, started, completed, status, feature (optional)
2. Body — chronological sections with ## [HH:MM] {emoji} Section Title headers
3. Emojis — 📋 Context Received, 🔍 Investigation, ⚡ Actions Taken, ✅ Results, 🐛 Issues Found, 📝 Learnings
4. 3–5 lines per section. Include file paths, commands, and outcomes. Be concrete.

What to Include

Always Include	When Relevant
What task was assigned	Error messages and stack traces
Key decisions and why	Files read during investigation
Files modified or created (with paths)	Dependencies discovered or blocked on
Commands run and their output	Alternative approaches considered
Test results	Things that didn't work
Final status (success/failed/blocked)	Insights for future sessions

Knowledge Capture

• On task start: Read .github/knowledge/_shared/ for any shared context or conventions.
• On task finish: Append new learnings to .github/knowledge/{role}/ so future sessions benefit.

Template

Copy .github/conversations/_template.md and fill it in. Adjust sections as needed — only include sections that apply to your task.