work

📋 Work: Plan Execution Engaged

Track tasks, manage plans, and survive CC compacts with three operating modes.

Activation

When this skill activates, output:

📋 Work: Plan execution engaged.

Then determine which mode to use based on the trigger.

Context Guard

Context	Status
User says "copy plan", "append plan", "resume plan"	ACTIVE: use matching mode
User says "what's next", "todo", "priorities"	ACTIVE: quick query mode
User provides a task list or plan	ACTIVE: copy mode
General discussion about planning concepts	DORMANT: do not activate
User is executing a task (not managing the list)	DORMANT: do not activate

Step 0: Silent Context Compilation (MANDATORY)

Before executing ANY mode, silently gather current state. Do NOT present findings. Do NOT ask questions. Just internalize:

1. Read the project's STATE.md (if it exists): current task, blockers, next steps
2. Read the project's CLAUDE.md (if it exists): conventions, architecture decisions
3. Check recent diary:

   python "$MEMSTACK_PATH/db/memstack-db.py" get-sessions <project> --limit 3
   

4. Check git state:

   git log --oneline -5
   git diff --stat
   

This is silent. Synthesize an internal understanding of where the project stands. Then proceed to the triggered mode with full context. The user already knows their project state, don't waste their time repeating it back.

Mode 1: Copy Plan

Trigger: "copy plan" or when a new plan is provided

1. Parse the entire plan into individual numbered tasks

2. For each task, save to SQLite:

   Write the payload to a file, then pipe it in with - as the argument. A payload on stdin is never seen by the shell's parser, so a redirection operator inside a task description cannot be read as one.

   cat task.json | python "$MEMSTACK_PATH/db/memstack-db.py" add-plan-task -
   

   task.json contains:

   {"project":"<name>","task_number":<n>,"description":"<task>","status":"pending"}
   

3. Confirm with task count

4. Also write a markdown copy to memory/projects/{project}-plan.md for human readability