Skills
Navigate
skills/hhhh124hhhh/skillmate/writing-style-coach/Gen Agent Trust Hub

writing-style-coach

Fail

Audited by Gen Agent Trust Hub on Sep 20, 2026

Risk Level: HIGHCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The handler function in scripts/style_learner.py accepts a user-controlled output_file path used in write operations without validation. \n
  • Evidence: The save_to_file function directly uses the output_file argument in open(output_file, 'w', encoding='utf-8'). An attacker can use path traversal sequences (e.g., ../../) to overwrite critical system files or shell profiles, enabling persistence or system compromise.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted article text which can contain malicious instructions intended to influence the AI agent during subsequent tasks.\n
  • Ingestion points: The articles parameter in scripts/style_learner.py.\n
  • Boundary markers: Absent. The script does not utilize delimiters or specific instructions to isolate sample content from the prompt generation logic.\n
  • Capability inventory: The skill includes file writing functionality (save_to_file in scripts/style_learner.py).\n
  • Sanitization: Absent. Extracted data from articles is interpolated directly into JSON and Markdown templates without escaping or instruction filtering.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Sep 20, 2026, 04:43 AM