← ClaudeAtlas

project-management-signal-extractionlisted

Extract project management signals from raw inputs (meeting transcripts, Slack threads, stand-up notes, async updates) into a structured signal map. Designed to run as a subagent so extraction doesn't blow up the orchestrator's context. This skill should be used when a project management skill or orchestrator needs to process raw signals before synthesis — never invoked directly by users.
matthull/my-skills · ★ 0 · AI & Automation · score 72
Install: claude install-skill matthull/my-skills
# Project Management Signal Extraction Extract structured project management signals from raw, messy inputs. This skill runs as a **subagent** — invoked by an orchestrating skill (e.g., `/project-management`), not directly by users. The orchestrator handles synthesis; this skill handles extraction with disciplined prompt rigor. Why a dedicated skill rather than an ad-hoc subagent prompt: extraction quality directly determines the accuracy of every downstream artifact (status updates, blocker reports, milestone tracking). An undisciplined extraction prompt produces wrong signals. This skill exists to make extraction correct and repeatable. --- ## What Is a Project Management Signal? A project management signal is any raw artifact that expresses the current state of work: where things stand, what's blocked, what was decided, what was committed to. Like product intent signals, they are expected to be partial, informal, and occasionally contradictory — that's normal. The extraction job is to surface the signal, not judge the source for being messy. **Signal types this skill extracts:** | Signal Type | Description | Example | |-------------|-------------|---------| | **Action item** | A commitment to do something, with owner and optionally a due date | "Tom will have the API spec done by Friday" | | **Status update** | Progress against a milestone or body of work | "We're about 70% through the indexing work" | | **Blocker** | Something actively preventing progress | "Can't