soia-dev-fix-loop
Pass
Audited by Gen Agent Trust Hub on Jul 20, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: No malicious patterns, obfuscation, or unauthorized remote code execution patterns were detected. The skill defines a procedural development workflow with defensive guidelines.
- [DATA_EXFILTRATION]: The skill includes explicit security boundaries that prevent the agent from outputting credentials, absolute paths, or private project metadata, serving as a safeguard against data exposure.
- [COMMAND_EXECUTION]: While the workflow involves executing commands for reproduction and testing, it is scoped to the local environment and includes instructions for the agent to seek user confirmation before performing high-risk or expansive operations.
- [INDIRECT_PROMPT_INJECTION]: The skill processes findings which may originate from external sources, representing an injection surface. Ingestion points: findings provided as text/table input. Boundary markers: normalization of input data. Capability inventory: file modification and shell command execution. Sanitization: the process relies on the normalization step to structure inputs before action.
Audit Metadata