Collab Compact
Productivitycollab-compactCollab Compact 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.
Save context and trigger compaction for clean resume
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/collab-compact/SKILL.md and all supporting files. Report provenance, permissions, scripts, dependencies, and security risks. Propose an exact installation into .agents/skills/collab-compact, 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.
Collab Compact
Save current collab session context and trigger compaction for a clean context resume.
When to Use
- Context is getting large and compaction is approaching
- Before a long break in the session
- Proactively to ensure clean state
Process
Step 1: Verify Active Session
ls -d .collab/*/ 2>/dev/null | xargs -I{} basename {}
If no sessions: "No active collab session. Use /collab first." STOP. If multiple sessions: Ask user which session.
Step 2: Save Context Snapshot
Read current state via MCP:
Tool: mcp__plugin_mermaid-collab_mermaid__get_session_state
Args: { "project": "<absolute-path-to-cwd>", "session": "<session-name>" }
Returns: { "phase": "...", "currentItem": ..., ... }
Determine activeSkill from phase:
- "brainstorming" → activeSkill = "brainstorming"
- "rough-draft/*" → activeSkill = "rough-draft"
- "implementation" → activeSkill = "executing-plans"
Save snapshot via MCP:
Tool: mcp__plugin_mermaid-collab_mermaid__save_snapshot
Args: {
"project": "<absolute-path-to-cwd>",
"session": "<session-name>",
"activeSkill": "<determined-skill>",
"currentStep": "<phase-from-state>",
"inProgressItem": <currentItem-from-state>,
"pendingQuestion": null,
"recentContext": []
}
Note: version and timestamp are automatically added by the MCP tool.
Step 3: Update State
Update collab state via MCP:
Tool: mcp__plugin_mermaid-collab_mermaid__update_session_state
Args: { "project": "<absolute-path-to-cwd>", "session": "<session-name>", "hasSnapshot": true }
Step 4: Trigger Compaction
Context snapshot saved to .collab/<session>/context-snapshot.json
Triggering compaction now...
Invoke the /compact command.
Step 5: Auto-Resume Session
After compaction, automatically resume the session:
Compaction complete. Resuming session...
Invoke skill: collab
This will restore context from the snapshot and continue where you left off.
Context Full Detection
When context usage is high before triggering compaction, render an Alert to notify the user:
Tool call:
Tool: mcp__plugin_mermaid-collab_mermaid__render_ui
Args: {
"project": "<absolute-path-to-cwd>",
"session": "<session-name>",
"ui": {
"type": "Alert",
"props": {
"type": "warning",
"title": "Context Full",
"message": "Run /compact in terminal, then /collab to resume."
}
},
"blocking": false
}
This provides a non-blocking visual notification without interrupting the skill execution.