Seal Verifier
Verify that an agent used a specific skill to produce a specific output. Paste a SkillSeal response to verify the cryptographic proof.
How Verification Works
1
Fetch Skill
Retrieve SKILL.md from GitHub at the claimed commit SHA
2
Hash Comparison
Verify the fetched content matches the seal's skill_text
3
Signature Check
Recover the EigenAI signer via EIP-191 and compare
The Trust Model
What the seal proves:
- The exact system prompt (skill) loaded during inference
- The exact output produced
- EigenAI's operator attests to both
What it doesn't prove:
- That the LLM perfectly followed the skill (LLM limitation)
- That the agent didn't run other un-sealed inferences
Security note:
Verifiers must fetch SKILL.md from the claimed commit independently. Never trust a cached or agent-supplied hash.