Add five clean-replay robotics task packages

#5

Adds NaVid, NaVILA, PoliFormer, SmolVLA-LIBERO, and TraceVLA-SimplerEnv to 12-robotics/. The public bundles implement the reviewed clean-replay evaluation path: delivery validation, source-only snapshot, generated-output cleanup, offline uid/gid-65532 replay, source manifest comparison, frozen evidence, and evaluate(EVIDENCE_ROOT).

The bundled v2 evidence policies describe the verifier-owned audit records and remaining evaluator limits. Public navigation ground truth is excluded. All five bundles passed static contract validation, JSON parsing, integrity verification, and a high-confidence credential scan; the clean source snapshot regression suite passed 34 tests.

Superseded by #8, which carries all 15 robotics task packages at the 2026-09-06 state plus verifier_refs_required.json. This branch predates this week's material fixes, and five partially overlapping robotics PRs made it impossible to tell which version was authoritative -- the same class of confusion that produced the image/package mismatch found by the 2026-09-04 runtime audit. Closing in favour of #8.

SueMintony changed pull request status to closed

Sign up or log in to comment