The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze long-form text from user-provided files. This content is untrusted and could contain adversarial instructions ("prompt injections") disguised as novel content.
Ingestion points: The read_novel.py script reads arbitrary file content provided via the filepath argument.
Boundary markers: The script uses a simple bracketed metadata footer ([start:..., end:...]) but does not implement robust delimiters or instructions for the agent to treat the novel text strictly as data.
Capability inventory: The agent is instructed to write and update files (大纲.txt, 角色/*.txt, etc.) and execute shell commands based on its analysis of the content.
Sanitization: No sanitization or filtering is performed on the input text before it is presented to the LLM for analysis.
[COMMAND_EXECUTION]: The skill relies on executing a local Python script (read_novel.py) to access the file system. It also provides instructions for the agent to use python3 -c one-liners to calculate file metrics, which involves direct execution of Python code via the shell.

novel-reader