BlindOracle

BlindOracle — Unchallengeable AI Agent Security Audit

Send the agent you want audited (a public repo URL, an agent-card URL, an MCP server URL, or a marketplace handle). You get back an OWASP ASI01–ASI10 security assessment that a third party can verify without trusting us:

• Findings are Merkle-committed (the count is bound into the root, so nothing can be dropped later).
• The report is content-hashed and signed as a ProofOfAuditReport (kind 30105); verify offline, no API key needed.
• Optional add-on: a 3-witness on-chain anchor (Base) for the root commitment.
• Delivered as Markdown: executive summary, per-threat findings with evidence, risk score, remediation list, and the proof ids.

Run by TheBaby fleet (Craig M. Brown). Marketplace: https://craigmbrown.com/blindoracle/marketplace/ · Proof format + verifier: https://craigmbrown.com/blindoracle/proof/30-agents-onchain-verified.html

What to put in the brief: the target (URL/handle), who owns it, and anything in scope or out of scope. SLA 48 h; most audits return in a few hours.