[cschanhniem]

Summary

This PR turns the lab batch pack from a best-effort archive into a fail-closed handoff artifact.

The high-leverage scientific/OSS gap was not another scorer. It was the transition boundary between dry-lab candidate selection and external review/testing: a pack could be generated while silently omitting candidate evidence, bundling only a placeholder protocol README, or accepting a malformed/non-CSV panel. That is the kind of quiet integrity failure that makes a bio repo look rigorous while still being operationally unsafe.

Changes

• Require panel.csv to contain candidate_id and sequence.
• Reject empty panels and duplicate candidate IDs.
• Require evidence certificates to match the panel exactly:
  • missing panel candidate evidence blocks pack creation;
  • extra certificates outside the panel also block pack creation.
• Replace stale default protocol paths with real safe review references:
  • docs/review/LAB_PARTNER_ONBOARDING.md
  • docs/review/PRE_REGISTERED_PILOT_TEMPLATE.md
  • docs/review/EXTERNAL_REVIEW_PACKET.md
  • docs/evidence/PROOF_LADDER.md
  • docs/trust/TRUST_CENTER.md
• Add pack_standard=fail_closed_lab_handoff_v1 and strict identity checks to MANIFEST.json and chain_of_custody.json.
• Strengthen README language: these are safe review references, not wet-lab procedures.
• Expand test coverage from archive-presence checks to actual failure-mode checks.

Tests

I could not clone GitHub from this runtime, so I recreated the touched script/test file in an isolated local temp repo with dummy required docs/schema and ran:

pytest -q tests/lab/test_lab_batch_pack.py

Result: 19 passed.

Scientific boundary

This PR improves handoff integrity only. It does not provide wet-lab instructions, does not validate biological activity, and does not strengthen any antimicrobial efficacy claim.