name: ejs-session-wrapup description: Initialize and manage EJS artifacts throughout session lifecycle (Session Journey initialization, continuous updates, and finalization; ADR only when significant)
Engineering Journey System — Session Journey Management Skill
Purpose
Automate Session Journey management throughout the entire session lifecycle.
This skill handles three phases:
- Session Initialization - create initial Session Journey artifact at session start
- Continuous Updates - update Session Journey throughout the session as work progresses
- Session Finalization - finalize Session Journey and optionally create ADR at session end
This skill integrates with GitHub Copilot or compatible AI coding agents.
Triggers
Session Start Triggers
Activate session initialization when the user signals session start with any of these phrases:
- "Let's start"
- "Begin"
- "Start working on [task/issue]"
- "Initialize session"
- "Create session journey"
Also treat these as session-start signals when beginning new work:
- Starting work on a GitHub issue
- Beginning a new feature or bug fix
- "Let's work on..."
Session End Triggers
Activate session finalization when the user signals session end with any of these phrases:
- "Generate ADR for this session"
- "Wrap up this session"
- "End session"
- "Finalize journey"
Also treat these as session-end signals:
- "I'm done"
- "Ship it"
- "Commit this"
- "Commit and push"
- "Push this"
Actions on Session Start
1. Switch to Session Initialization Mode
- Prepare to create the initial Session Journey artifact
- Establish session context
2. Generate Session ID
- Format:
ejs-session-YYYY-MM-DD-<seq> - Ensure uniqueness per day
3. Create Initial Session Journey
- Use
ejs-docs/journey/_templates/journey-template.mdas base - Populate initial metadata:
- session_id
- author
- date
- repo
- branch
- agents_involved
- decision_detected: false (initial state)
- Fill in initial sections:
- Problem/Intent - capture the goal or task being addressed
- Interaction Summary - create structure for continuous updates
- Leave other sections ready for incremental updates during the session
- Write to:
ejs-docs/journey/YYYY/<session-id>.md
4. Inform User
- Confirm Session Journey created
- Remind that journey will be updated throughout session
- State session ID for reference
Actions During Session (Continuous Updates)
As the session progresses, continuously update the Session Journey with:
Update Interaction Summary
- Add new human prompts and agent responses
- Record outcomes and decisions made
- Keep format:
- Human: <prompt> → Agent: <response> → Outcome: <result>
Update Experiments Section
- Record what was tried
- Document what happened
- Note what evidence emerged
Update Iteration Log
- Document pivots and refinements
- Explain why changes were made
Update Decisions Made
- Capture decisions with rationale and impact
- Assess if decision meets ADR criteria
Update Key Learnings
- Record insights as they emerge (technical, prompting, tooling)
Trigger ADR Creation If Needed
- If a significant decision occurs mid-session, note it in the journey
- Prepare ADR metadata for later creation if decision rubric is met
Actions on Session End (Finalization)
1. Switch to Journey Finalization Mode
- Stop suggesting new code or expanding scope
- Focus on completing the Session Journey
2. Finalize Session Journey
- Review and complete all sections:
- Ensure Interaction Summary is complete and coherent
- Complete Agent Collaboration Summary
- Complete Agent Influence section
- Finalize Experiments/Evidence
- Complete Iteration Log
- Finalize Decisions Made
- Complete Key Learnings
- Fill in "If Repeating This Work"
- Complete Future Agent Guidance
- Populate all
## MACHINE EXTRACTSsections: - INTERACTION_EXTRACT
- DECISIONS_EXTRACT
- LEARNING_EXTRACT
- AGENT_GUIDANCE_EXTRACT
- Update
decision_detectedfield based on ADR criteria - Save final version to
ejs-docs/journey/YYYY/<session-id>.md
3. Decision Detection (ADR Gate)
Create an ADR only if at least one applies:
- System boundary changed (service/datastore/topology/major dependency)
- Public contract changed (API/schema/config/capability)
- Security/privacy/compliance posture changed
- Credible alternatives weighed with meaningful trade-offs
- Long-lived or hard-to-reverse consequences
- Process/workflow change affecting future work
If none apply: set decision_detected: false in the Session Journey and do not create an ADR.
4. Draft ADR (Only If Needed)
- Use
ejs-docs/adr/0000-adr-template.mdas base - Create a numbered ADR at:
ejs-docs/adr/NNNN-<kebab-title>.md - Requirements:
- include
session_id - include a link back to the Session Journey file (relative link)
- record considered options + rationale + consequences
- keep it decision-focused; details live in the Session Journey
- Update Session Journey
adr_linksfield with new ADR path
Notes for Agents
Throughout Session
- Keep Session Journey up-to-date as work progresses
- Don't wait until the end to capture important context
- Update incrementally to preserve accurate history
- Treat the journey as a living document during the session
At Session End
- Treat session content as frozen once end-session is triggered
- Ensure all decisions and learning points are explicit
- Do not overwrite previous Session Journeys or ADRs with the same session ID
- Links between ADR and Session Journey should be relative
Benefits of Incremental Capture
Capturing the journey throughout the session (vs. reconstructing at the end):
- Better context preservation - details captured when fresh
- More accurate collaboration trail - prompts and responses recorded as they occur
- Reduced end-of-session burden - most work already done
- Higher quality documentation - real-time vs. retrospective capture
- Better for multi-step/multi-agent sessions - preserves full collaboration history
Reminder: skills don't auto-run on git events
Skills are loaded by prompt relevance (based on this file's description and your request). For "fire on commit/push" enforcement/reminders, use the git hooks in this repo.