Problem
The final required Fable High review for #175 certified the candidate but reported a non-blocking P2: tests/test-release.sh rejects any literal ADAPTER_TEST_ONLY occurrence in tests/test-adapter.sh. That proves the current canonical entrypoint has no inherited focus selector, but it is grep-fragile: a harmless comment or documentation line could fail the release suite.
The check was added as a bounded correction after both required reviewers found that an inherited focused-test selector could make a canonical proof falsely green. The security invariant is valid; the literal-text implementation deserves a narrow follow-up rather than another mutation of the already certified #175 candidate.
Source checkpoint: #175 (comment)
Acceptance criteria
- Preserve the invariant that the canonical proof cannot inherit a focused adapter-test selector or silently skip adapter coverage.
- Replace the raw literal grep with a behavioral test or a clearly scoped proof-runner/test-entrypoint contract.
- Harmless comments and documentation must not trigger the guard.
- Keep focused development commands, if any, explicitly outside the canonical proof path.
- Add TDD RED then GREEN coverage.
- Run focused proof, canonical 11/11 proof, self-review, and normal Sol High + Fable High review.
- Synchronize ROADMAP.md when this issue becomes active.
Release placement
Non-blocking follow-up from the #175 certifying review. Schedule for 1.1.0 harness hardening; it does not reopen or weaken the 1.0 gate.
Problem
The final required Fable High review for #175 certified the candidate but reported a non-blocking P2: tests/test-release.sh rejects any literal ADAPTER_TEST_ONLY occurrence in tests/test-adapter.sh. That proves the current canonical entrypoint has no inherited focus selector, but it is grep-fragile: a harmless comment or documentation line could fail the release suite.
The check was added as a bounded correction after both required reviewers found that an inherited focused-test selector could make a canonical proof falsely green. The security invariant is valid; the literal-text implementation deserves a narrow follow-up rather than another mutation of the already certified #175 candidate.
Source checkpoint: #175 (comment)
Acceptance criteria
Release placement
Non-blocking follow-up from the #175 certifying review. Schedule for 1.1.0 harness hardening; it does not reopen or weaken the 1.0 gate.