Verifier Intent Coverage — mechanically guarantee the contract covers scope intent
Make the contract's fidelity to scope intent **mechanically trustless**. Today the contract is positioned as the authoritative specification, but it is only a *fallible proxy* for what the scope actually asked for. Whether built code satisfies intent rests on the planner's diligence translating intent→assertions and the verifier's disposition to *notice* a thin translation — with **no mechanical check that every acceptance criterion is even covered, and no gating outlet when a gap is found.**
verdict PASSscore 35 / 35findings 10 (0 risk · 3 debt · 7 obs)duration 278h 58mrejection cycles 0shipped Jun 16, 2026surface cli
Pipeline timeline
Intent to proven code in 278h 58m across Think, Plan, and 2 Build→Verify phases.
Think16634m
Plan22m
Build 122m
Verify 120m
Build 229m
Verify 28m
Assertion ledger
35 claims, each independently verified. Showing 8 — show all →
| ID | Says | Matcher | |
|---|---|---|---|
| A001 | Acceptance criteria written as a dash-bullet are recognized | verified | ok |
| A002 | Acceptance criteria written as a heading are recognized | verified | ok |
| A003 | Acceptance criteria written in bold are recognized | verified | ok |
| A004 | Acceptance criteria written as a bare label are recognized | verified | ok |
| A005 | A well-formed scope is never mistaken for an unreadable one | verified | ok |
| A006 | Across every past project scope, none is wrongly judged unreadable | verified | ok |
| A007 | Across every past project scope, the criteria are always recovered | verified | ok |
| A008 | A scope in an unfamiliar format is flagged as unreadable, not guessed | verified | ok |
Findings 10 total
debt→ scope
Flaky test in the broader suite — one test failed in the sealed verify run (3766p/1f/2s) but never reproduced across 7 full-suite runs + 5 regression-focus runs. 'Push failed after retry' noise points to a git-operation/retry test, not Phase 1 code.
obspackages/cli/tests/commands/scope-ac-corpus.test.ts→ monitor
scope-ac-corpus.test.ts asserts toBe(0) against the live, growing completed-scope corpus (205 today). A future completed scope using '## Acceptance Criteria' with non-AC-id criteria would flip emptyExtractionCount/falseAmbiguousCount and break this test in an unrelated future PR. Intentional safety gate, but couples future greens to historical scope formatting.
obspackages/cli/src/commands/artifact-validators.ts→ monitor
Bold-form regex /\*\*\s*(AC\d+)/ matches **ACn** anywhere on a line, so a prose mention ('see **AC3** above') extracts AC3 as a criterion id. Harmless across the 205-scope corpus and the dogfood (exactly AC1-14, no spurious ids), but a future version-1.1 contract could gain a spurious prose-derived id that becomes a false 'uncovered' block. Latent, low-likelihood.
obspackages/cli/src/commands/artifact-validators.ts→ acknowledge
Defensive try/catch around joinCoverage in evaluateCoverageGate is effectively unreachable — joinCoverage is total (no throw path), so the 'could not evaluate' diagnostic branch cannot trigger in practice and no test exercises it. Spec explicitly requested defensive depth, so it is intentional, but it is untested dead-ish code.
debtpackages/cli/tests/utils/proofSummary.test.ts→ monitor
Stale/cross-contract @ana tags in long-lived test files mis-map this contract's assertion IDs
+5more findings
Provenance
Who ran what, and what it cost. Recomputable estimates from the shared price table — subordinate to the verdict, never gating.
model claude-opus-4-8
| session | turns | tools | in | out | cache | cost |
|---|---|---|---|---|---|---|
| ana | 107 | 44 | 8.4k | 90.0k | 5.3M | $5.98 |
| plan | 94 | 41 | 7.7k | 70.5k | 4.8M | $5.23 |
| build | 102 | 47 | 7.7k | 51.9k | 4.1M | $4.13 |
| verify | 104 | 45 | 8.6k | 33.1k | 3.5M | $3.24 |
| build 2 | 314 | 166 | 12.0k | 87.9k | 21.7M | $14.66 |
| verify 2 | 90 | 42 | 7.6k | 27.3k | 2.2M | $2.40 |
| TOTAL 6 sessions | $35.63 (table 2026-06-14) | |||||
churn 26 files · +1813/−38
completeness ✓ complete (plan 1/1 · build 2/2 · verify 2/2)
Integrity seal
scopesha256:299e88902e431...
contractsha256:5095b2cec3666...
plansha256:769fd994ce674...
spec-1sha256:d5f1665657541...
spec-2sha256:665f7fdad48da...
build-report-1sha256:8aa8795f15fd6...
build-data-1sha256:dad91d60d5219...
verify-report-1sha256:53b1a8d8bf67e...
verify-data-1sha256:1cfe055ceeefd...
build-report-2sha256:ed865254234ce...
build-data-2sha256:02a9f1987306b...
verify-report-2sha256:3d1f744977093...
verify-data-2sha256:18ccd430e582a...
audit cmd$ ana proof audit verifier-intent-coverage → all hashes match