Proof Gaps

This page records known missing or weak proof as first-class evidence state. Where the gap is a concrete missing artifact, the page uses an expected-failing Specdown check so the report stays visibly red without making the suite unusable.

How To Read A Gap

A proof gap means a promise or cross-cutting rule has an open evidence condition. The Evidence Gaps concern is the product rule that weak evidence stays visible; this page is the current ledger of those open proof conditions. Some gaps are next implementation work; others are waiting for an explicit policy choice or an expensive proof run. Expected-failing checks keep concrete missing artifacts visible in the report.

Known Gaps

gap id	affects	missing proof	unlock trigger	close condition	status
`gap.traceability-config`	model, all views	Specdown trace graph is not configured for typed promise/shared-concern/proof documents.	decide trace types and edge cardinality	`specdown trace -strict` passes without expected failure	open
`gap.improve-held-out-cycle`	`promise.improvement`, `rule.cost-and-proof-freshness`	No checked-in end-to-end improve cycle with held-out validation.	run or fixture a representative bounded improve cycle	durable held-out cycle summary is linked from evidence map	open
`gap.vocabulary-evidence-bundle`	`rule.vocabulary-consistency`	No dedicated vocabulary consistency evidence bundle across prose, JSON, tests, and agent guidance.	choose deterministic vocabulary probe scope	current vocabulary evidence packet exists	open
`gap.live-batch-fixture`	`promise.evaluation`, `rule.host-owned-execution`	Live invocation and batch invocation proof are still partial.	create host-owned runtime fixture with sanitized packets	live/batch fixture output is linked from evidence map	open
`gap.review-learning-active-run-aggregation`	`rule.agent-human-resumability`	`review feedback summarize` can count explicitly selected `cautilus.review_feedback.v1` packets, but no active-run or report surface discovers the packet set across runs.	decide active-run/default packet location and packet discovery rules	an active-run or report route selects review-feedback packets, runs the summary, and links the durable summary artifact from the evidence map	open

Expected-Failing Proof Checks

gap.traceability-config

specdown trace -strict

specdown: no trace configuration found in config

run:shell

gap.improve-held-out-cycle

test -f artifacts/self-dogfood/improve-held-out-cycle/latest/eval-summary.json

exit status 1

run:shell

gap.vocabulary-evidence-bundle

test -f .cautilus/claims/evidence-vocabulary-consistency-current.json

exit status 1

run:shell

gap.live-batch-fixture

test -f artifacts/self-dogfood/live-batch/latest/result-packet.json

exit status 1

run:shell