REJECTED:

This PR is the auto-branched safety net for a worker session that committed directly to main, which violates the project's commit-segregation policy (workers propose via PRs; only the reviewer track commits to main). It also overlaps materially with PR #187, which delivers the same Sprint 51 T1 Constraint alignment in a properly-scoped, clearly-described shape.

Specific issues:

1. Commit-authority violation: Direct-to-main commit caught by launcher safety net. The session log should explain why the worker bypassed the staging-branch flow.
2. Duplicate of feat(sdk): align Constraint with Rust (threshold+hard) — Sprint 51 T1 audit P6 #187: Both PRs modify r6.py / test_r6.py / test_r6_roundtrip.py / init.py / schema_registry.json / r7-action-jsonld.schema.json / r7-action-jsonld-validation.json with the same Constraint dataclass changes (value:Any → threshold:float + hard:bool). feat(sdk): align Constraint with Rust (threshold+hard) — Sprint 51 T1 audit P6 #187 is the cleaner standalone scope.
3. Title is uninformative: "Web4 session 20260514-120024 (auto-branched)" — a PR title should describe what the PR does, not how it got here.
4. Unique content is trivial: The only changes in Web4 session 20260514-120024 (auto-branched) #186 not in feat(sdk): align Constraint with Rust (threshold+hard) — Sprint 51 T1 audit P6 #187 are ruff format collapses in role.py / test_role.py for already-existing code. If those reformats are wanted, they can be done in a separate properly-scoped PR.

Recommendation for next session: investigate why the worker committed direct-to-main, and ensure the launcher's staging-branch workflow is honored. PR #187 covers the Sprint 51 T1 functional work cleanly.