acl-submission

Pass

Audited by Gen Agent Trust Hub on Sep 23, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill consists of purely instructional markdown content designed to guide an AI agent through a manual review process for academic submissions. It does not contain executable code, script invocations, or network requests to external domains.
  • [INDIRECT_PROMPT_INJECTION]: The skill establishes an attack surface by directing the agent to process untrusted external data, such as paper PDFs, supplemental archives, and repository metadata, for auditing purposes.
  • Ingestion points: The skill expects the agent to review submission materials (PDFs, repositories, supplements).
  • Boundary markers: None specified within the skill instructions.
  • Capability inventory: The skill itself does not invoke any system tools or network operations.
  • Sanitization: Not present, as the skill provides natural language instructions for a human-in-the-loop or agent audit rather than automated data parsing.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 23, 2026, 11:28 PM
Security Audit — agent-trust-hub — acl-submission