reproduce-bug-report
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from GitHub issues, support reports, and tickets to guide the execution of child agents with high-privilege computer-use capabilities.
- Ingestion points: External bug report data is extracted in the 'Parent workflow' and subsequently passed into the 'Shared child prompt' in SKILL.md.
- Boundary markers: The skill lacks explicit boundary markers or 'ignore' instructions for embedded content when interpolating untrusted text into the agent prompts.
- Capability inventory: Child agents are granted significant control over a cloud desktop environment via the
remote.computer_use_enabled: trueconfiguration, including file system and UI interaction. - Sanitization: No programmatic sanitization or escaping of external content is performed; the skill relies on the parent agent to 'extract' information before passing it to the child agents.
Audit Metadata