Fetch Product Context

Fetch Product Context

This is a utility skill that loads product context from the consumer repo's data/ directory. Other skills call this pattern before doing their work.

What this skill reads:

• data/decisions/ — commitments the PM has made
• data/signals/active.md — recent time-stamped observations
• data/knowledge/ — people, reference, research, market-landscape
• data/personas/ — defined customer personas
• data/tasks/active.md — active backlog and in-progress work

See the DB Routing Rubric in .claude/context/data-schemas.md for what each entity type holds.

How to Identify the Current Product

1. Read the repo's CLAUDE.md file
2. Look for the "Repo Identity" section
3. Extract the product name
4. If no Repo Identity section exists, ask the user which product they're working on. (One product per repo: the repo is the product.)

How to Read Data Files

The consumer repo follows the layout in .claude/context/data-schemas.md. Always read indexes before opening individual files — this keeps token usage bounded.

Decisions (always fetch)

1. Read data/decisions/index.md (one-line-per-decision table).
2. Filter rows by Status = Active and Date within the last 90 days (extend if thin). Keep all Status = Active regardless of age if the decision is type Positioning, Pricing, or Architecture (these are long-lived constraints).
3. Open the 3–10 most relevant decision files referenced from the index.
4. Summarize: what was decided, when, and any constraints imposed.

If data/decisions/ does not exist, note it: "No decisions logged yet for this product."

Personas (fetch when skill needs user context)

1. Read data/personas/index.md.
2. Open the primary persona file (or all if multiple are clearly distinct).
3. Extract: who they are, JTBD, pain, evidence strength.

If data/personas/ is empty, suggest running /define-persona.

Backlog priorities (fetch when skill needs scope context)

1. Read data/tasks/active.md.
2. Extract the top 10 items by priority order (Now → Next → Later H2 sections, in order).
3. Note current phase from the host CLAUDE.md if specified (Explore, Validate, Build, Scale).

Recent Signals (fetch when skill needs market or user-feedback context)

1. Grep data/signals/active.md for H3 headings with date: metadata within the last 30 days (extend to 60 if thin).
2. Group by type from the metadata: User Feedback, Technical Constraint, Market Signal, Competitive Move, Internal Learning.
3. Highlight any with action_required:true.
4. If looking for evaluation context, also note any decisions of type Scope or Go-to-Market from the decisions step.

Market Landscape (fetch when skill needs competitive context)

1. Glob data/knowledge/market-landscape/*.md.
2. Pick the file matching the current product's market category (the file name is typically {market-slug}.md).
3. Read only the most recent ## Scan — {date} H2 section, not the full file history.
4. If the latest scan is older than 30 days, note it and suggest running /market-scan to refresh.

If no Market Landscape file exists, note it: "No prior scans on file — suggest running /market-scan."

Output

Present a brief context summary to the user:

• Product: [name]
• Key decisions: [2-3 bullet summary, cite by file path]
• Persona: [one-line summary]
• Current focus: [phase + top priorities]
• Recent signals: [count by type, highlight any action-required]
• Market scan freshness: [date of last scan, or "none on file"]

This skill can be invoked directly for a quick context briefing, or used as a foundation step by other skills.