The Agent Skills Directory

[PROMPT_INJECTION]: The skill exposes an indirect prompt injection surface by ingesting and processing untrusted user input (theme descriptions) to influence agent actions and prompt generation.
Ingestion points: User theme descriptions are analyzed and extracted in SKILL.md (e.g., the '理解用户输入' section).
Boundary markers: Absent. The instructions do not specify the use of delimiters or 'ignore' instructions for user-provided data.
Capability inventory: The agent is instructed to execute local Python scripts (generator_engine.py and narrative_prompt_generator.py) using arguments extracted from user input.
Sanitization: Absent. There is no mention of sanitizing, escaping, or validating the theme strings before they are used as parameters in Python code blocks.
[COMMAND_EXECUTION]: The instructions explicitly direct the agent to generate and execute Python code snippets to interact with local generation engines. User-provided strings for themes and styles are passed directly as method arguments (e.g., theme, style_keywords), which represents a potential injection vector if the underlying scripts do not implement strict input validation.

prompt-generator