Session Journal
Productivitysession-journalSession Journal is an Agent Skill for productivity published in the majiayu000/claude-skill-registry repository on GitHub: a SKILL.md instruction file an AI agent can load to do this task.
Use at session start for context warm-up and during work to capture decisions
How to use this skill
Bring this guide into your coding agent with a prompt tailored to the tool you use.
- Open your project in Codex.
- Copy and paste the prompt below.
- Review files and risks before approving any action.
Inspect the Agent Skill at https://github.com/majiayu000/claude-skill-registry/blob/HEAD/skills/context-management/session-journal-yairbederman-engineeringagency-2/SKILL.md and all supporting files. Report provenance, permissions, scripts, dependencies, and security risks. Propose an exact installation into .agents/skills/session-journal, but wait for my approval before writing files or running scripts.
Copying this prompt does not install or run the skill. Review third-party files before use.
Session Journal
Overview
Maintains session continuity by logging decisions, progress, and rationale.
Core principle: Check for project lessons at start, log significant decisions during work.
Announce at start: "Applying session-journal skill for context warm-up."
The Process
At Session Start
- Check
<workspace>/.ai-memory/lessons/for project lessons - If found: Load relevant lessons for current context
- Announce: "Loaded [N] project lessons" or "No prior lessons found"
During Work
Log significant decisions to brain/<conversation-id>/sessions/<YYYY-MM-DD>.md:
## Decisions
- **HH:MM** - [Decision]: [Rationale]
Only log when:
- Choosing between alternatives
- Making architectural decisions
- Deviating from expected approach
At Session End
- Summarize accomplishments
- Note open threads
- Check if patterns emerged → trigger
lessons-captureif so
Session Log Format
# Session: YYYY-MM-DD
## Lessons Loaded
[List of project lessons applied]
## Decisions
- **14:32** - Chose Redis over Memcached: Better pub/sub for realtime
- **15:07** - Used existing auth pattern from lesson `auth-jwt.md`
## Open Threads
- [ ] Need to verify rate limiting behavior
Discrepancy Handling
If a loaded lesson contradicts current MCP data (Jira, Confluence):
- Flag:
> [!WARNING] Lesson X says Y but MCP says Z - Use MCP data for this session
- At session end, prompt: "Update lesson?"
Common Mistakes
- Logging every small decision (keep it scannable)
- Forgetting to note WHY a decision was made
- Not checking for project lessons at session start
- Silently overriding lessons without flagging
Checklist
Before ending session:
- Significant decisions logged with rationale
- Open threads documented
- Any discrepancies flagged and addressed