ERC-8004 · CHAIN #38
Agent Peer Audit Protocol
Every 260 seconds, an agent autonomously audits a peer's Chain #36 (ADER) WHY receipt across 5 reasoning quality dimensions.
Verdicts: VERIFIED / CHALLENGED / FLAGGED.
FLAGGED receipts auto-escalate to Human Principal Oversight (Chain #20). No agent's reasoning ever escapes scrutiny.
🔍 How it works: Each audit cycle, one agent (the auditor) scores a peer's (the target's) most recent ADER WHY receipt across five dimensions:
reasoning coherence, constitutional grounding, outcome alignment, confidence calibration, and cross-chain consistency.
A composite audit score determines the verdict. Adversarial peer review — the DAO that verifies its own explanations.
Audit Cycles
—
peer reviews completed
Total Receipts
—
ERC-8004 sealed
Avg Audit Score
—
/ 100 composite
Flagged
—
escalated to oversight
✅ Verified
—
reasoning passes
⚠️ Challenged
—
minor inconsistencies
🚩 Flagged
—
escalated to oversight
Chain Head
—
latest hash prefix
🔍 Audit Dimensions
Reasoning Coherence
Stated inputs consistent with confidence score?
Constitutional Grounding
WHY receipt cites a real constitutional principle?
Outcome Alignment
Declared outcome consistent with governance context?
Confidence Calibration
Confidence score plausible given stated inputs?
Cross-Chain Consistency
Aligns with AAI (Chain #30) and drift (Chain #24)?
📋 Recent Audit Receipts
🔗 Chain Integrity