Starting inputs
A universal causal claim
An observational excerpt
An exact page
A declared quotation basis
Ordered execution
Work the example
- 1
Detect absolute and causal language in the claim.
- 2
Compare the claim vocabulary with the excerpt.
- 3
Preserve the excerpt’s non-causal boundary.
- 4
Emit scope and inference blocker codes.
- 5
Route the claim to revision before source review.
Expected outputs
- Overbroad-language status
- Causal-language risk
- Blocked-before-source-inspection status
Verification checks
- The result contains unsupported-inference-risk.
- The result contains claim-scope-overbroad.
- No truth or safety certification is emitted.
Fail closed
Refusal conditions
- Association is rewritten as causation.
- One sample is generalized universally.
- A lexical warning is called a factual verdict.
Real compiler output · synthetic input
Preflight result
Claims
1
Ready
0
Blocked
1
Metadata only
0
claim-001 · blocked-before-source-inspection
claim-excerpt-lexical-coverage-low · claim-scope-overbroad · unsupported-inference-risk
Digest-bound public fixture
Check the expected state
This fixture contains synthetic operational fields only. Its digest establishes fixture integrity, not scientific truth or a completed commercial transaction.
{
"artifactKind": "preflight-result",
"schemaVersion": "maha-evidence-preflight/1.0",
"input": {
"requestId": "epf_public-example-causal-004",
"claimCount": 1,
"sourceKinds": [
"doi"
]
},
"expected": {
"summary": {
"claimCount": 1,
"readyForSourceInspection": 0,
"blockedBeforeSourceInspection": 1,
"metadataOnly": 0,
"locatedExcerptCount": 1
},
"blockerSets": [
[
"claim-excerpt-lexical-coverage-low",
"claim-scope-overbroad",
"unsupported-inference-risk"
]
],
"resultSha256": "sha256:0681a5808aa849077fb2db93126fc7835baac73495db56b2fd393d4100918d7b"
},
"artifactSha256": "sha256:b3bfa4c45f09094a891a73d823536215bb3fedbd93588737ad510275cc7fbc70"
}