| name | evidence-research |
| description | Use when sources need to be converted into a claim ledger with confidence levels. |
evidence-research
Lifecycle stage: RESEARCH
Trigger
Use when external claims, markets, users, or technical options need verification.
When not to use
Do not use when this trigger is absent; choose the command or skill that owns the requested state, artifact, and verification gate.
Inputs
- User request or current artifact.
- Known constraints and context.
- Relevant evidence or source links, if available.
- Current Agent Brain state.
Procedure
- State the current state and target artifact.
- Identify missing blockers and ask at most three blocking questions.
- Separate facts, assumptions, hypotheses, and open questions.
- Apply the anti-rationalization table below.
- Produce the required artifact: Research Claim Ledger.
- Add evidence, risks, decision, and next state.
Anti-Rationalization
| Shortcut | Rebuttal |
|---|
| "This is obvious." | Write the assumption and evidence. If you cannot, it is not obvious. |
| "We can do this later." | If the missing step changes the decision, do it now or state the risk. |
| "The user wants speed." | Reduce scope; do not skip the quality bar. |
| "This does not need verification." | Every important claim or behavior needs proof. |
Verification
- Required artifact exists and is named.
- Facts, assumptions, and open questions are separated.
- Evidence or evidence gaps are explicit.
- Next state is stated.
- Stop conditions are honored.
Output Artifact
Research Claim Ledger
Use templates/research-claim-ledger.md. The artifact should be concise, auditable, and include evidence, blockers, and next action so another agent can resume.
Failure Modes
- Producing advice instead of an artifact.
- Accepting user assumptions without challenge.
- Skipping evidence because the task feels simple.
- Recommending an agent when a simpler system is enough.
- Hiding risks or open questions.
Example
Trigger: a claim needs source-backed evidence before planning or public copy. Action: split the claim into subclaims, inspect canonical sources, mark confidence, and reject unsupported adoption language. Output artifact: templates/research-claim-ledger.md with blockers and next action. Verification: cite source provenance, checked scope, confidence, and recheck trigger.
A claim says, "Developers already want this workflow." The skill should split the claim into source-backed subclaims, record canonical sources, mark unsupported adoption language as unknown, and produce a research claim ledger with recheck triggers before any roadmap or public copy uses the claim.
