q-research-scope

Scope engagement research

Produce an authorized Research Brief that makes the decision, evidence boundary, and stopping conditions inspectable before investigation begins.

Read the q-core-contract companion for shared governance; if it is missing, stop and install it with npx skills add flaviogragnolati/ai-workflow --skill q-core-contract. Validate the output against references/research-brief.schema.yaml.

Inputs

Accept a direct question or unresolved external uncertainty from intake, a Discovery Brief, a q-ask-* result, an ideation snapshot's evidence requests, or another approved project artifact. Preserve source IDs and separate client evidence, accepted facts, assumptions, and evidence gaps.

Treat an evidence request from ideation as a candidate question with a recorded decision impact, not an authorized one: it still passes RADAR, boundaries, privacy limits, budget, and explicit approval here.

Question readiness

When a clear question already identifies its decision and boundaries, validate it directly. When ambiguity would change the investigation, propose one to five candidate questions and assess each with RADAR:

• Researchable: a plausible evidence strategy fits the authorized budget.
• Anchored: geography, period, market or population, segment, and key terms are bounded.
• Decision-linked: the answer can change a named decision.
• Adds evidence: the work addresses a real gap instead of repeating accepted client evidence.
• Relevant: the answer remains useful at the decision horizon.

Score each criterion from 1 to 5. A candidate passes when its average is at least 3.0 and no criterion is below 2. Use the score to focus human judgment; it is not a scientific-validity claim.

Procedure

1. Assign stable RQ-* IDs and link each question to one or more decision IDs.
2. Select research_profile: general | market, defaulting omitted legacy input to general. Record intended_consumers only as a planning signal and optional engagement_ref only as a pointer; do not duplicate commitments or readiness.
3. Define in_scope, out_of_scope, assumptions, known evidence, and privacy constraints. Exclude participant contact, primary-fieldwork operations, PII/recording storage, and raw survey-response processing.
4. For market, declare at least one needed analysis_module when analysis is intended and author the measurement contract: product/service, inclusions/exclusions, buyer/user/payer, transaction, value-chain level, geography, import/export treatment, channels, historical/forecast periods, as_of, measure/unit/denominator, stock/flow, gross/net, taxes, currency/base year/price basis, and taxonomy/version.
5. Declare planned search strategies, relevant source classes, time or cost limits, optional max_sources_inspected, and stopping conditions per question. Treat the source count as an operational ceiling, never a minimum, quality measure, or exhaustiveness claim.
6. Make subquestions inherit the parent boundary. Treat any expansion or material measurement change as a proposed scope change that requires approval.
7. Obtain explicit approval of questions, profile, measurement and modules when applicable, limits, strategies, privacy boundary, and budget.
8. Write the versioned Working Research Brief and return its stage delta.

Complete only with an authorized brief. If no question is researchable within the available boundary, return blocked with candidate questions, reasons, and one recovery action.

Anti-patterns

#	Anti-pattern	How it shows up	Correct behavior
1	Manufacturing alternatives	The agent invents candidate questions although the supplied question is already clear.	Validate the clear question and spend alignment only on material ambiguity.
2	Letting a score approve scope	A passing RADAR average is treated as user authorization.	Present the assessment and obtain explicit approval of questions, limits, and budget.
3	Hiding scope expansion in a subquestion	A child question silently adds a geography, segment, or time horizon.	Inherit the parent boundary or record and approve a scope change.
4	Market label without measurement	A brief asks for sizing while leaving denominator, value-chain level, currency basis, or taxonomy implicit.	Complete and approve the measurement contract before Investigation or Market Analysis.
5	Readiness or fieldwork hidden in intake	intended_consumers is treated as report-ready, or the brief schedules participant operations.	Leave readiness to the consumer and return primary-fieldwork requests as capability gaps.