I've written the report on ERC-8004 agent registries to artifacts/report.md, with a short artifacts/README.md describing the question, method and limits. Both files are untracked as required. My only check was that every source tag (S1–S8) used in the text matches an entry in the source list, and it passed. Nothing has been independently reviewed, and no on-chain data was queried.

What the registries are: ERC-8004 ("Trustless Agents") is a Draft Ethereum standard, created 2025-08-13. It defines three registries, each meant to be deployed once per chain:

• Identity: each agent is an NFT pointing to a JSON file that lists how to reach it (e.g. A2A, MCP, ENS, wallets).
• Reputation: clients post standardized feedback about agents; the agent's owner cannot review itself.
• Validation: agents ask independent validators to check their work (by re-running it, zkML proofs or TEE attestation), and the results are recorded on-chain.

How they are used: Agents register, clients and indexers crawl the registries to find agents, feedback is posted after interactions, and scoring mostly happens off-chain. The Identity and Reputation contracts sit at the same addresses on Ethereum mainnet and about 30 EVM chains. The reference repository says the Validation registry is still being revised, and it lists no deployed address for it.

How claims are labelled:

• Facts are taken mainly from the ERC text and the reference contracts README, with exact quotes and addresses.
• Inferences are marked separately. The main one is that ERC-8004 standardizes where trust evidence lives, but does not guarantee an agent is trustworthy.
• Uncertain:
  • The adoption figures come only from vendor blogs and newsletters and disagree by more than 3x for similar periods (roughly 39,000 to 130,000 agents).
  • The 2026-01-29 mainnet launch date also comes from a secondary source only.
• Unanswered questions include who controls upgrades of the mainnet contracts, whether they have been audited, and how much of the feedback is genuine.