Proposal (rationale)

Installing @rungs/cli@0.3.1 with the tracked profile into a clean clone of Arena Lab registered 21 gates, then failed immediately: findings-disposition-has-reason interpreted the scaffold’s own | — | | | *nothing open* | | | | placeholder as a finding with missing Sev, Pri and Evidence. A generated baseline that fails its own schema makes first use start with a waiver and prevents the consumer from distinguishing its defects from Rungs’ defects.

Decision

accepted — 2026-09-05, by the user as the first bootstrap repair. Fix the general empty-register case in Rungs before adopting the tracked package in Arena Lab.

Plan

Requirements

• The findings register emitted by the module passes findings-disposition-has-reason while it contains no real finding.
• A malformed real open row still fails with the missing required fields named.
• The behavior is generic and does not special-case Arena paths or content.
• A fresh tracked scaffold reaches the check runner without this failure.

Impacts

• src/engines2.ts register-schema parsing or the findings module’s generated empty state.
• Focused engine tests and a packaged/scratch tracked-scaffold regression.
• Module version and upgrade behavior only if emitted content changes.

Approach

First reproduce the smallest table input and trace whether placeholder rows should be omitted by the producer or recognized as empty by the schema engine. Prefer one generic representation of an empty register. Preserve enforcement for any row that claims a real identifier.

Acceptance criteria / tests

1. A focused test reproduces the 0.3.1 failure before the fix.
2. The generated empty findings register produces zero schema findings after the fix.
3. A real row missing Sev, Pri or Evidence still produces the expected findings.
4. A fresh tracked install followed by rungs check has zero failures attributable to the empty findings register.
5. npm test and this repository’s rungs check pass without weakening an existing gate.

Out of scope

• Implementing builders for every currently named but unrun self-test fixture; those remain visible coverage debt and are not the cause of this false failure.
• Existing-authority migration, CLI pinning, ref-only Git behavior or publishing; WI-066 through WI-068 and a later release item own those concerns.

Execution

Implemented on feature/WI-065-fresh-findings-baseline from 084802e.

• Taught the generic register-schema engine that an em dash in the identity column represents an explicit no-record row, even when another cell carries explanatory placeholder text.
• Added an engine-level regression proving the placeholder is ignored while a real finding still reports missing Sev, Pri and Evidence.
• Added the generated placeholder as a passing module self-test.
• Added a package-level journey that initializes a clean tracked Git repository and immediately runs rungs check successfully.

Review

Verified on 2026-09-05.

1. Met. The focused engine test reproduces the 0.3.1 sentinel shape and covers the corrected result.
2. Met. The generated empty row produces zero findings and zero examined records.
3. Met. A real F-001 row still reports empty Sev, Pri and Evidence fields.
4. Met. The package-level test initializes a clean tracked Git repository and its immediate rungs check exits zero with the findings gate present.
5. Met. npm test passes 38/38 and npm run rungs -- check passes 29/29 registered gates. The existing transparent warning that 45 fixtures have no builder remains and is out of scope.