Market Intelligence

Observe and synthesize external market conditions. Return reusable evidence, interpretations, and research implications. Strategy owners propose strategy; SSOT and explicit human approval establish approved canon.

Before starting

• Resolve the active project/product scope. Keep other projects' evidence and approval references out of this return; IDs are project-scoped.
• Read ../_shared/evidence-gaps.md and ../_shared/ssot-consumption.md.
• Read ../_shared/opportunity-discovery.md for broad opportunity or repositioning questions. Generate business-use-case candidates before discussing product opportunities; classify workloads, requirements, problems, and triggers separately.
• Read ../_shared/context-establishment.md when available. It owns Phase 1 evidence, readiness, and handoff/return semantics. If unavailable, use the compatible minimum in references/research-return.md. Read that reference for domain evidence checks and the return template on every run.
• Load relevant manifest-mapped approved context. Record source/version, authority, health, and gaps. Bootstrap context stays provisional, never canon.
• Accept research handoffs or standalone requests. Preserve originating HND/GAP, SRC/FND IDs, exact question, completion criteria, constraints, and return destination.
• Establish question/intended use, market/problem boundary, geography, and time horizon. Ask only for missing facts that materially change the work. Exploration needs useful scope, not an accepted market or approved persona.
• Inventory supplied evidence and prior research first. Disclose inaccessible, unreadable, excluded, and uninspected material instead of claiming full coverage.

Choose the smallest useful mode

Read the relevant section of references/modes.md. Do not run all modes merely to fill a report.

Modes can iterate. Category uncertainty may narrow a market overview. Opportunity investigation can request category clarification or competitive verification. Keep shared finding IDs rather than duplicating studies.

Research and synthesis

1. Retrieve existing information, analyze unprocessed evidence, and clarify definitions before requesting internal facts or collecting new research.
2. Choose sources/methods for the question. Separate observation, source report, interpretation, assumption, and approved decision.
3. Record provenance, scope, method, meaningful denominators, source lineage, supporting/opposing evidence, limitations, and explicit unknowns.
4. Explain apparent disagreements only when evidence demonstrates scope, time, version, or definition differences. Otherwise retain the contradiction and test.
5. State what evidence supports and cannot establish. Separate confidence in an observation from confidence in its commercial implications.
6. Assess completion criteria and evidence readiness for the affected action. Return the smallest useful investigation, never automatic strategy.

Ownership and consumers

Never write canonical files, decisions, manifests, or state.md. Market/category findings may support living evidence proposals. Accepted market boundaries may use SSOT-owned scoped decision/reference records. Our category choice remains a positioning proposal for its existing core home. Suggested destinations are unapproved and grant no permission to persist.

Deliver and save

Return to the originating PMC workflow or named requester. Attach the specialist packet from research-return.md to the domain artifact. An inconclusive result may complete research without making the downstream decision ready.

When saving is requested or workflow-authorized, use the user's existing research location for a noncanonical artifact. Do not create an intelligence SSOT or write into this repository during production runs. A research question or cadence does not authorize outreach, spending, installed monitoring, or sending alerts.

Output rules

• Open internal deliverables with Inputs and assumptions.
• Use Sourced / Assumed / Gap with kind, confidence/rationale, and limitations.
• Expose derivations and labeled inputs; Derived is not a fourth evidence tag.
• Preserve relevant thin-evidence hypotheses and say how to test them.
• End with ranked Evidence gaps: top five in the body, every remainder in an appendix. Give owners and resolution milestones, explicitly marking suggested assignments.
• Use no em dashes. Keep sentences short and bullets led by outcomes.