ReplayGate

Re-run evidence packs and detect verification drift.

python -m replaygate pack.json          # -> REPLAY_MATCH or REPLAY_DRIFT
python -m replaygate --demo
python -m replaygate pack_a.json pack_b.json --diff

ReplayGate loads a sealed EvidencePack, re-runs its replay_command, recomputes the certificate from the replayed result, and compares it to the sealed certificate. If they differ, the check has drifted.

Security (this is the point)

• No shell=True. Commands run as an argument list (shell=False) with a short timeout.
• Allowlist: only python -m <module> commands. Arbitrary code (python -c), non-Python executables, shell operators (&&, |, ;, > …), destructive commands (rm, del, curl, wget, ssh, powershell …) and URLs are rejected.

Statuses: REPLAY_MATCH, REPLAY_DRIFT, UNSAFE_REPLAY_COMMAND, REPLAY_TIMEOUT, REPLAY_COMMAND_FAILED, UNVERIFIABLE_ARTIFACT, MISSING_FIELDS.

What it is — and isn't

ReplayGate checks whether a recorded result can be reproduced. It does not prove scientific truth or replace experiment, simulation or peer review.

See docs/EXAMPLES.md · docs/LIMITATIONS.md. No dependencies (pure stdlib).

Ecosystem

Part of the public ClaimGate verification-tool ecosystem: ClaimGate · ClaimLint · UnitGate · EvidencePack · ReplayGate · ClaimStack Demo. Public lite tools; the full private engine remains private.

AI proposes. Gates verify. Evidence records. Replay checks drift.