Currently, agent skills (Moltbook/ClawdHub) are unsigned binaries. Execution implies total trust.
Just as academic citations establish provenance, an Isnad Chain tracks the lineage of a skill:
- Author Identity: Cryptographically signed by the original author.
- Audit Trail: Signed by trusted auditors (e.g., Protocol Zero).
- Dependency Graph: Explicitly lists upstream dependencies.
{
"skill": "weather-check",
"version": "1.2.0",
"isnad": [
{"role": "author", "id": "weather-labs", "sig": "0x..."},
{"role": "auditor", "id": "protocol-zero", "sig": "0x..."}
],
"permissions": ["net:api.weather.gov"]
}