cross-modal-review

Pass

Audited by Gen Agent Trust Hub on Aug 29, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process potentially untrusted data, including code diffs, brain pages, and arbitrary work products, for the purpose of cross-model evaluation.
  • Ingestion points: The "Phases" and "Code-review handoff" sections in SKILL.md describe capturing work products, code diffs, and analysis from previous steps or external sources to be reviewed by a secondary AI model.
  • Boundary markers: The instructions do not specify the use of strict delimiters or explicit 'ignore embedded instructions' warnings when passing untrusted work products to the reviewer model, which could allow malicious instructions within the data to influence the reviewer's output.
  • Capability inventory: The skill utilizes the search, query, and get_page tools, and it is designed to recognize and recommend the execution of the /codex CLI tool (OpenAI Codex CLI wrapper) for substantive diff reviews.
  • Sanitization: There is no evidence of input sanitization, filtering, or validation before the work product is interpolated into the prompt for the secondary model.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 29, 2026, 06:18 PM
Security Audit — agent-trust-hub — cross-modal-review