chinese-novelist

Warn

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: MEDIUMCOMMAND_EXECUTION
Full Analysis
  • COMMAND_EXECUTION (MEDIUM): The skill instructs the agent to execute a Python script (scripts/check_chapter_wordcount.py) using a command-line argument that includes the user-defined novel name (novels/[小说名称]/...). This creates a vulnerability where a malicious user could provide a novel name containing shell metacharacters (e.g., ;, |, or `) to execute arbitrary commands if the agent's execution environment does not perform strict sanitization.
  • INDIRECT_PROMPT_INJECTION (LOW): The skill has a surface for indirect prompt injection by reading from user-provided inputs and generated files to drive subsequent logic.
  • Ingestion points: Phase 1 user responses and the generated 00-大纲.md file.
  • Boundary markers: Absent; the skill does not use delimiters or warnings to ignore instructions within the ingested data.
  • Capability inventory: File system access (reading and writing in novels/) and local Python script execution.
  • Sanitization: None; the skill does not describe any validation or escaping of the user-provided story details or filenames.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 17, 2026, 06:45 PM