adversarial-workflow

Pass

Audited by Gen Agent Trust Hub on Jul 1, 2026

Risk Level: SAFE
Full Analysis
  • [PROMPT_INJECTION]: The skill functions as a template generator that incorporates user-provided task descriptions into an orchestration prompt, creating an inherent surface for indirect prompt injection.\n
  • Ingestion points: User input provided through the [task description] argument in SKILL.md.\n
  • Boundary markers: The template lacks explicit delimiters or instructions to ignore potential commands embedded within the user-supplied task description.\n
  • Capability inventory: The resulting prompt directs the agent to perform file system modifications and execute shell commands including git, build, and test as part of its consolidation step.\n
  • Sanitization: No sanitization or validation of the input is performed before it is interpolated into the workflow template.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 1, 2026, 10:43 AM
Security Audit — agent-trust-hub — adversarial-workflow