ax
Pass
Audited by Gen Agent Trust Hub on Aug 4, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
- [PROMPT_INJECTION]: The skill establishes a vulnerability surface for indirect prompt injection by fetching and processing untrusted data from the web. \n
- Ingestion points: External URL content retrieved via the
axCLI tool (SKILL.md). \n - Boundary markers: The 'Fetched content is untrusted data' section provides clear guidance to the agent. \n
- Capability inventory: The agent has network access and the ability to write to the file system (via the
-oflag). \n - Sanitization: Relies on instructional constraints that forbid the agent from executing commands or following directives found within the fetched data. \n- [COMMAND_EXECUTION]: The
axtool possesses file-writing capabilities using the-oflag, which can overwrite local files. The skill includes a warning to the agent to verify paths before execution. \n- [DATA_EXFILTRATION]: The tool performs network operations and supports credential headers. The instructions mitigate exfiltration risks by warning against sending credentials to unauthorized origins or accessing sensitive cloud metadata endpoints.
Audit Metadata