The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill operates by ingesting user-provided story ideas, outlines, and full-text drafts to perform automated expansion, diagnosis, and polishing, which exposes the agent to instructions embedded within untrusted user data.
Ingestion points: The skill takes 'core梗' (core plots) in references/chapter4-ai-workflow.md and full novel text in references/chapter5-polish.md to perform '去AI味' (de-AI flavoring) processing.
Boundary markers: There are no specified delimiters or 'ignore' instructions provided to the agent to distinguish between the user's creative writing and potentially malicious instructions embedded within that writing.
Capability inventory: The agent has the capability to generate content, call external workflows, and provide financial/tax guidance based on input.
Sanitization: The instructions do not define any sanitization or validation protocols for the text provided for polishing or expansion.
[EXTERNAL_DOWNLOADS]: The skill encourages the use of an external, third-party AI writing platform that is not recognized as a well-known service or a verified vendor resource.
Evidence: Files references/chapter4-ai-workflow.md and references/chapter6-monetize.md direct users to use https://xingyuexiezuo.com/?workflow=1906 for one-click content generation and workflow execution.

fanqie-masterclass