run-qc
Pass
Audited by Gen Agent Trust Hub on Sep 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input from
QC_<date>.mdfiles and translates them into actions, creating a surface where malicious instructions in the plan could influence agent behavior. - Ingestion points: Reads and executes items from local
QC_<date>.mdfiles. - Boundary markers: The skill explicitly instructs the agent to 'Prove the boundary' and verify that credentials are 'absent, not merely forbidden,' providing a logical security check.
- Capability inventory: Access to browser automation, system logs, and shell command execution.
- Sanitization: No specific technical sanitization (e.g., escaping or delimiter wrapping) is mentioned for the content of the QC plan before it is processed by tools.
- [COMMAND_EXECUTION]: The skill instructions explicitly direct the agent to 'run commands' and 'drive the product' as part of the QC process, which is a powerful capability that requires careful oversight when driven by external documentation.
Audit Metadata