file-format

Pass

Audited by Gen Agent Trust Hub on May 14, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: Indirect Prompt Injection Surface. The skill is designed to read formatting rules from resources/FORMAT.md and allows users to update these rules via the 'Updating Standards' instructions. This creates a persistence mechanism where malicious instructions could be stored in the formatting file and subsequently executed when the agent retrieves rules for specific document types or purposes.
  • Ingestion points: The agent reads content from resources/FORMAT.md using grep search patterns (SKILL.md).
  • Boundary markers: While the skill uses markdown headings (##) and horizontal rules (---) to segment the data, it lacks explicit 'ignore embedded instructions' warnings for the agent when processing this content.
  • Capability inventory: The skill can read/write local files and directly influences the context for downstream file-generation tools (PDF, DOCX, PPTX generators).
  • Sanitization: There is no evidence of validation or sanitization when the agent updates the formatting standards file based on user requests.
  • [COMMAND_EXECUTION]: The skill instructs the agent to perform local command-line operations to retrieve data.
  • Evidence: SKILL.md explicitly directs the agent to "Use grep to search resources/FORMAT.md for the section keyword" to extract formatting rules.
Audit Metadata
Risk Level
SAFE
Analyzed
May 14, 2026, 12:14 AM
Security Audit — agent-trust-hub — file-format