qemu-agent-feedback
Pass
Audited by Gen Agent Trust Hub on Jul 31, 2026
Risk Level: SAFE
Full Analysis
- [DATA_EXFILTRATION]: The skill is designed to send feedback to GitHub. This potential risk is mitigated by a mandatory sanitization process that replaces sensitive information with placeholders and requires the user to approve the final content and target repository before the gh issue create command is executed.
- [COMMAND_EXECUTION]: Uses standard system tools (git, gh) for workspace auditing and issue filing. The commands are structured safely, using file-based inputs rather than direct shell interpolation of potentially untrusted data.
- [PROMPT_INJECTION]: The skill handles workspace logs and commands as input, representing an indirect prompt injection surface. The instructions mitigate this by requiring the agent to avoid raw logs and instead provide high-level behavioral summaries. Evidence: (1) Ingestion points: workspace artifacts and logs in SKILL.md; (2) Boundary markers: sanitization table and boundary instructions in SKILL.md; (3) Capability inventory: gh issue create and git commands in SKILL.md; (4) Sanitization: comprehensive replacement rules for secrets, paths, and credentials defined in SKILL.md.
Audit Metadata