ln-601-semantic-content-auditor

Fail

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: CRITICALPROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • PROMPT_INJECTION (LOW): The skill is susceptible to Indirect Prompt Injection (Category 8) because it ingests and processes the content of untrusted project documents.
  • Ingestion points: Project documents loaded via the doc_path variable (e.g., architecture.md, api_spec.md).
  • Boundary markers: The skill looks for a specific SCOPE comment but does not employ delimiters or 'ignore' instructions for the rest of the document body.
  • Capability inventory: Access to Bash, Read, Grep, and Glob tools with broad filesystem visibility.
  • Sanitization: None; document content is processed directly to generate scores and findings.
  • COMMAND_EXECUTION (LOW): During the verification of runbook.md, the instructions tell the agent to 'test command syntax'. This creates a risk where the agent might execute arbitrary shell commands provided within an audited document, potentially leading to unauthorized system actions if the document is compromised.
  • EXTERNAL_DOWNLOADS (SAFE): The diagram.html file includes a script tag pointing to cdn.jsdelivr.net for the Mermaid library. While this is an external dependency from a non-whitelisted source, it is a standard library used solely for visual representation of the workflow and does not impact the agent's core execution environment.
Recommendations
  • Contains 1 malicious URL(s) - DO NOT USE
Audit Metadata
Risk Level
CRITICAL
Analyzed
Feb 17, 2026, 06:07 PM