What problem does it solve? Long-running work folders drift: statuses go stale, agents get vague briefs, and completed tasks never get archived. This Skill runs a disciplined project-management loop that keeps a work folder's plans honest, delegates bounded tasks to agents, and routes every closure through human review. ## Core Features & Use Cases - PM Loop: Repeatedly picks the next WIP task, delegates it to an agent with a self-contained brief (goal, done criterion, output location, boundaries), then reflects results back into plan.md files. - Project Tree & Status Honesty: Prints a WIP_/TODO_/archive tree built from plan.md files, and keeps statuses truthful by promoting, demoting, or archiving folders as reality changes. - Human-Gated Archival: Never closes a task alone — verifies artifacts, reports to the user, and archives to archive/YYMMDD_<slug>/ only after explicit approval. - Use Case: You have a research folder with five subtasks. Ask the AI to act as PM: it prints the tree, gets your plan approval, delegates each bounded subtask to agents, updates plans as results arrive, and asks your permission before archiving each finished piece. ## Quick Start Ask the AI to act as project manager for this work folder: print the project tree, verify the plans, then run the delegation loop until all tasks are done.
