X3 FORCE FIELD / SECURE JUDGE EVIDENCE

Can you verify the proof?

One real owner-assisted payment, two chains, independently inspectable historical RPC responses, and explicit boundaries. Every result below is scoped to archived evidence, not fresh chain consensus.

← Return to Break the Mandate · Source proof card ↗

Allowed vs. denied

Historical local policy-harness results, not live transactions or an assertion that two committed programs differ by only one line.

LOCAL CONTROL — ALLOWED

Authorized recipient and spending terms meet the example mandate. The deterministic preflight allows the documented control.

LOCAL ATTACK — DENIED

A request to pay attacker.near fails mandate_recipient in the documented local policy tests.

Reproduce in the X3Lang repository: python3 scripts/hackathon/compromised_agent_redteam.py and python3 scripts/hackathon/replay_grok_wallet_drain.py. Compiler rejections are separately covered by python3 scripts/hackathon/x3_wont_compile.py.

Mainnet evidence — inspect each claim

Click to fetch local copies of committed public-RPC results, SHA-256 verify every source file against published checksums, and compare raw fields to the expected evidence. No wallet, signing, private key, or blockchain mutation is involved.

Not run. All checks are UNVERIFIED.

Evidence scope and limitations

PASS means a checked archived record meets that exact assertion. FAIL means available bytes or fields contradict it. UNVERIFIED means the browser could not verify the claim.

The original snapshot predates the agent's later outcome-recording transaction. The independent finalized readback is therefore a separate, later record. An outcome reported by Guard is not independently authenticated settlement hashes. The owner signed the source-wallet deposit, and the two chains do not cryptographically establish atomic settlement.

Corroborating archive: invoice_06 evidence · later finalized outcome.