The Agent Skills Directory

[PROMPT_INJECTION]: The skill includes explicit instructions to bypass safety guidelines regarding sensitive content. SKILL.md states that fictional elements like domestic violence, betrayal, or violence must be extracted and that the agent 'must not refuse entire chapters or books on the grounds of sensitivity or discomfort.'
[COMMAND_EXECUTION]: The skill employs local command execution for data processing. SKILL.md contains a bash pipeline using ls, sed, and sort to merge chapter summaries into a single file. Additionally, references/style-profile-generator.md dynamically generates and executes a Python script to perform sentence length and punctuation density statistics on sampled text.
[PROMPT_INJECTION]: Indirect Prompt Injection Surface. The skill processes untrusted external data (novel text) and has significant local capabilities.
Ingestion points: Reads raw novel content from user-provided file paths or direct text input.
Boundary markers: While it uses a [材料声明] (Material Declaration) prefix to provide context to sub-agents, it does not implement robust delimiters to prevent embedded instructions in the novel text from influencing the agent's behavior.
Capability inventory: The skill can write numerous files to the local filesystem (拆文库/{书名}/) and execute shell and Python commands.
Sanitization: No explicit sanitization or escaping is performed on the ingested novel text before it is processed by the analysis pipeline.

story-long-analyze