Known traps and negative knowledge

2 min readReferenceUpdated 2026-05-21

What you'll find here

Two related contracts agents use to avoid repeating known mistakes: craik.known_trap (evidence-backed pitfalls) and craik.negative_knowledge (evidence-backed statements about what is not true or not available in a scope).

Absence needs evidence too.

Use negative knowledge only when absence matters and there is evidence for it — inspected repository tree, fresh tool attestation, etc. Don't promote unsupported guesses about absence into negative knowledge.

Contracts

Contract

Records

Lifecycle

craik.known_trap

pitfall

Trap statement · avoidance guidance · evidence · handoff links · optional expiry · whether the trap is active, expired, or contradicted.

craik.negative_knowledge

absence

Bounded negative statement · scope · trust class · evidence · handoff links · contradictions · optional expiry.

Active, unexpired traps and negative knowledge are surfaced.

Onboarding and case-file stale-risk warnings include active known traps and active negative knowledge so agents can avoid repeating known mistakes or relying on disproven availability claims.

Capture commands

Command

Writes

Boundary

craik knowledge trap

known trap

Requires an active operator session and evidence ids. The record guides future agents but does not grant policy authority.

craik knowledge negative

negative knowledge

Requires evidence and scope. If a positive assertion is supplied as contradicted, Craik opens a contradiction instead of silently replacing it.

What's next

ReferenceKnown traps viewThe operator surface formatter for traps.ReferenceFreshnessHow knowledge expiry interacts with stale-risk warnings.GuideAgent onboardingThe first surface that exposes active traps.