compare-approaches

Pass

Audited by Gen Agent Trust Hub on Sep 1, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill establishes a process for analyzing technical designs that involves ingesting untrusted data from the repository environment, which presents a surface for indirect prompt injection.
  • Ingestion points: The agent is instructed to cite repository evidence for assumptions and evaluate the current design by inspecting the system (SKILL.md).
  • Boundary markers: The instructions do not specify the use of delimiters or provide explicit warnings to the agent to ignore instructions embedded within the technical documents or code it analyzes.
  • Capability inventory: The skill explicitly mentions that the agent should 'build or run the smallest isolated probe' to resolve uncertainties (SKILL.md), indicating the use of the host environment's execution capabilities.
  • Sanitization: The instructions do not include steps for sanitizing or escaping content retrieved from the repository before processing it for decision-making.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 1, 2026, 06:02 PM
Security Audit — agent-trust-hub — compare-approaches