CI's empty corpus is a named skip, and the documents catch up to the …

[akminx]

…build

tests/test_api.py probed /funds at import time. CI applies the migrations to an empty Postgres - the corpus is gitignored - so periods[0] raised IndexError during collection and 946 tests never ran: eight consecutive red pushes while the local gate stayed green against the loaded demo schema. The probe is now a session-scoped fixture, and the six ledger-dependent tests skip with a reason SANCTIONED_SKIPS names, exactly as the four other corpus-dependent suites already do. Seeding CI a ledger rich enough to satisfy 0 < unsupported < total means maintaining a second corpus shaped to produce specific verdicts - the fixture-written-to-fit failure this project has already had once. Verified: 20/20 against demo, 14 passed + 6 sanctioned skips against a freshly migrated schema, mypy at ceiling, 966 collect.

The README, Blueprint and roadmap described the system as designed rather than as built. Model extraction has shipped - Lucra's CEO email fired the stated trigger and the citation binding accepted three of five figures - retrieval is built to layer 2 and measured both ways (40/40 scoped, 24/40 blind), settlement gates R1, and the ledger holds 253 cited facts. The scope note now leads with what the gates cannot see, and each new cut carries its trigger - including a relationship graph projected over the ledger once multi-hop questions on the firm's real data stop being writable as SQL.

Summary

Risk tier

• routine
• semantic
• trust-critical

Path-derived minimum (if known): ___________

Oracle / evidence

• Spec section(s):
• Oracle path (tests/goldens/properties):
• Command run + result:

Review topology (from review-policy.yaml)

• Deterministic gate green (check_all / CI)
• Bugbot (mechanical only — not semantic approval)
• Pass A — semantic adversary
• Pass B — implementation adversary (required for semantic+)
• Pass A / Pass B / Bugbot each ran blind — no pass was shown another's findings
• Finding-owner verified fixer changes (if any)
• Human Judge (required for trust-critical)
• CODEOWNERS approval (trust-critical paths — n/a on solo repos, see below)

Fresh-eyes check (the solo substitute for code-owner review)

• I read this diff in a separate session from the one that produced it
• For trust-critical: I read the diff itself, not just the summary of it

Implementer model family

Unresolved assumptions

Test plan

• [ ]

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
7gc-os	Ready	Preview, Comment	Jul 28, 2026 8:20am