Run the outer agentic loop once. Triage the backlog, pick the most valuable ready item, shape its spec, and hand it to the inner loop. Use when the user asks "what's next", "groom the backlog", "run the PM", or on a schedule. Pass --dispatch to label the chosen issue for CI instead of building locally.
Scanned 9/4/2026
Install to Claude Code
npx -y skills add chrille0313/agentic-monorepo-template --skill pm --agent claude-codeInstalls into .claude/skills of the current project.
Are you the author of Pm?
Add the live security badge to your README — it updates automatically with every re-scan.
[](https://www.skillsdirectory.com/skills/chrille0313-agentic-monorepo-template)More formats (shields.io, HTML) on the badges page.
---
name: pm
description: Run the outer agentic loop once. Triage the backlog, pick the most valuable ready item, shape its spec, and hand it to the inner loop. Use when the user asks "what's next", "groom the backlog", "run the PM", or on a schedule. Pass --dispatch to label the chosen issue for CI instead of building locally.
---
# /pm: the outer loop (orchestrator)
Input (`$ARGUMENTS`): optionally `--dispatch`.
## 1. Triage
Spawn the `pm` agent for a triage. Its deliverables: backlog health (stale items, blocked items, items missing acceptance criteria), the ONE most valuable ready-or-readyable item, and a build-ready spec for it.
## 2. Report
Relay to the user: the pick and why, the spec, plus a short health note (e.g. "3 items lack acceptance criteria: #12, #15, #18").
## 3. Hand off
- **Without `--dispatch`** (interactive): ask the user whether to run `/build` on the spec now. If yes, invoke the build skill with the spec.
- **With `--dispatch`** (autonomous/CI): post the spec as a comment on the chosen issue, then add the `agent` label, which triggers `.github/workflows/agent-task.yml` to run the inner loop headlessly. Independent tasks run in parallel: after dispatching, repeat the triage while ready items remain that are independent of everything now in flight (each pm pass picks knowing what the previous one dispatched). Two limits are real, and only these two: a dispatch must not depend on or touch the same code as in-flight work, and the human merge gate paces everything: when unreviewed agent PRs are piling up, stop dispatching rather than adding more.
If nothing on the backlog is actionable, say so, list what would make the top items ready, and stop. An empty dispatch is a fine outcome.
Is this your skill, or is something wrong with this listing? . Author removals are honored within 72 hours.
No comments yet. Be the first to comment!