qa

Pass

Audited by Gen Agent Trust Hub on Jul 27, 2026

Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The agent uses the gh issue create tool to interact with GitHub. User input is processed and formatted into issue titles and bodies which are passed to this command.
  • [DATA_EXFILTRATION]: The skill transmits data to an external service (GitHub). To protect sensitive information, the instructions explicitly forbid including file paths, line numbers, or internal implementation details in the public issues.
  • [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection (Category 8) due to the processing of untrusted user data.
  • Ingestion points: User-provided bug descriptions (Interactive session).
  • Boundary markers: Absent; user input is interpolated into issue templates and subsequently into gh command arguments without explicit delimitation.
  • Capability inventory: GitHub CLI usage (gh issue create) and background codebase exploration (SKILL.md).
  • Sanitization: The agent is instructed to follow specific Markdown templates and use project-specific domain language (UBIQUITOUS_LANGUAGE.md), which serves as an instruction-based filter for the external input.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 27, 2026, 10:25 AM
Security Audit — agent-trust-hub — qa