The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted text from the user or session history and inserts it directly into a command argument template, creating an injection risk.
Ingestion points: User input text following /webnovel-learn or text from the current conversation is ingested as described in SKILL.md.
Boundary markers: No boundary markers or instructions to escape special characters are present to isolate the data from the command context.
Capability inventory: Uses the Bash tool to execute a Python script (webnovel.py) which updates project files.
Sanitization: Lacks any character escaping or validation mechanisms for the {用户输入或提炼后的完整描述} placeholder, which could lead to syntax breakdown or shell parameter manipulation if the text contains quotes, backticks, or subshell expansions.
[COMMAND_EXECUTION]: The skill relies on executing local bash commands to invoke a Python script (webnovel.py) within the project workspace. Although this is standard for its primary feature set, executing commands with dynamically constructed arguments from external input poses inherent security risks if the input is malicious.

webnovel-learn