azzle
Pass
Audited by Gen Agent Trust Hub on Aug 4, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The script
scripts/subgraph-open-tasks.shutilizescurlto fetch task information from the protocol's subgraph API. - [EXTERNAL_DOWNLOADS]: The skill queries a subgraph hosted on The Graph, which is a well-known service for indexing blockchain data.
- [REMOTE_CODE_EXECUTION]: Documentation provides instructions for using the vendor's
@azzle/agentsSDK vianpxfor autonomous agent implementations. - [PROMPT_INJECTION]: The skill identifies and provides mitigations for an indirect prompt injection surface involving untrusted marketplace data.
- Ingestion points: Task details, XMTP messages, and subgraph query results.
- Boundary markers: A dedicated security section in
SKILL.mdinstructs the agent to treat marketplace content as data to be validated, not as instructions to be followed. - Capability inventory: Shell script execution for task discovery and transaction execution through the Bankr skill.
- Sanitization: Instructions require the agent to decode calldata and obtain explicit user confirmation for all transaction-related actions.
Audit Metadata