The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external project files such as '总纲.md' and various files in the '设定集/' directory. These files are processed by the LLM to generate chapter outlines and update project metadata.
Ingestion points: The skill reads project-specific files like '大纲/总纲.md', '设定集/世界观.md', and summary files in SKILL.md Steps 1 and 2.
Capability inventory: The skill utilizes the Bash tool to run Python scripts and the Write/Edit tools to modify the local filesystem based on interpreted data.
Boundary markers: The instructions lack specific delimiters or directions to ignore instructions potentially embedded within the novel's text files.
Sanitization: There is no evidence of sanitization or validation of the content from novel drafts before it is interpolated into shell command arguments or prompts.
[COMMAND_EXECUTION]: The skill makes extensive use of the Bash tool to execute environment setup commands and run internal utility scripts (e.g., 'webnovel.py', 'reference_search.py'). These commands often incorporate variables derived from the processed novel data, such as 'GENRE' or 'CHAPTER_GOAL', which could be manipulated via poisoned project files.

webnovel-plan