troubleshoot

Pass

Audited by Gen Agent Trust Hub on Mar 3, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it incorporates untrusted user input (the 'issue' argument) directly into a diagnostic and remediation workflow.
  • Ingestion points: The issue description and error messages provided as arguments in the '/troubleshoot' command defined in SKILL.md.
  • Boundary markers: No specific delimiters or instructions (like 'treat the following text strictly as data') are used to separate the user-provided issue description from the agent's logic.
  • Capability inventory: The skill documentation indicates it leverages Bash for runtime diagnostics and has a '--fix' flag to automatically apply changes to the system or codebase.
  • Sanitization: There is no evidence of input validation or sanitization to ensure that instructions embedded within the 'issue' text are not executed by the agent.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 3, 2026, 08:06 AM