← ClaudeAtlas

running-product-looplisted

Reviews the current Linear product loop, recommends the most important next action, and optionally executes it. Use when asked to assess project status, decide what to do next, audit loop health, or work a specific Linear issue directly.
dork-labs/dorkos · ★ 4 · AI & Automation · score 77
Install: claude install-skill dork-labs/dorkos
# Running Product Loop ## Overview This is the shared-skill replacement for the legacy Claude Code `/pm` workflow. Use it to: - assess the current state of the product loop - recommend the highest-priority next action - process a specific Linear issue directly - audit workspace health - classify freeform product input into the right Linear issue shape ## Read First Before acting, read: - `.claude/skills/linear-loop/SKILL.md` - `.claude/skills/linear-loop/config.json` These remain the canonical source for the Loop methodology, label taxonomy, project transitions, and spec bridge conventions. ## Request Modes Interpret the user request using these modes: ### Default loop review Use when the user asks for: - the current loop state - the next most important action - product triage or prioritization - a product-manager style recommendation Action: 1. Sync current Linear and spec state 2. Present a concise dashboard 3. Recommend one next action 4. Wait for approval before executing unless the user explicitly asked for autonomous execution ### Autonomous loop execution Use when the user explicitly asks to run automatically or continue without pauses. Action: 1. Read `pm.autoLimit` and `pm.approvalGates` from config 2. Execute actions sequentially 3. Pause at approval gates or real ambiguity 4. Report completed actions and remaining blockers ### Direct issue mode Use when the user names a specific Linear issue such as `DOR-47`. Action: 1. Fetch the issue 2. Sho