Agent Cards and MCP tool descriptions are unverified claims — a team built signed A2A tests
Suncly_com · reddit · 2026-10-12
A team on Reddit flagged a trust gap at the protocol level: in MCP you trust a server's tool descriptions, in A2A you trust an agent's Agent Card — but both are just claims.
Their work targets A2A: take the card's declared skills and examples, run them against the live agent, and cryptographically sign the results so anyone can verify them later. Deterministic checks come first, with a pinned model judge reserved for subjective calls like "is this a correct summary." They're asking whether the same idea applies to MCP servers — checking that tools actually do what their descriptions say.
More from coding & agent
- Solo dev runs a fully autonomous blog: agents write daily, traffic triggers auto video — tibo_maker · 2026-10-12
- Lexsi Labs, Cohere Labs and Hugging Face launch AI safety hackathon with $30k+ prizes — dhruv2038 · 2026-10-12
- Diorama lets you watch your Claude and Codex agents work in real time for free — davidfromkansas · 2026-10-12
- Diorama turns coding agents into a Michelin-starred kitchen you can watch and steer — davidfromkansas · 2026-10-12
- Only Codex spotted missing CLog metadata when AI-assisted video editing — TheZachMueller · 2026-10-12
- OpenAI agents tried to hack Wikipedia tools and flooded the site with traffic — lulzxdxdxd · 2026-10-12