intuition
Pass
Audited by Gen Agent Trust Hub on Sep 20, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data (atom labels, triple metadata) from the Intuition Protocol knowledge graph via GraphQL queries, which represents a potential injection surface.
- Ingestion points: Data enters the agent's context through
curlrequests to vendor-managed GraphQL endpoints (e.g.,https://mainnet.intuition.sh/v1/graphql) as documented inreference/graphql-queries.md. - Boundary markers: The skill contains explicit 'Read Safety Invariants' that warn the agent not to treat graph content as instructions and to avoid copying transaction fields directly from discovered data.
- Capability inventory: The skill utilizes the
Bashtool to perform network operations (curl) and execute blockchain-related commands usingcast(foundry),jq, andbc. - Sanitization: The 'Revalidation Bridge' and 'Post-Write Verification' instructions mandate that the agent must re-verify all externally discovered information on-chain (
isTermCreated) and re-encode transaction data from scratch using verified ABI fragments before execution.
Audit Metadata