The Agent Skills Directory

[SAFE]: The skill manages novel projects by reading and writing files within the local workspace. No unauthorized data access or external network operations were detected.
[SAFE]: The skill correctly attributes its resources to the vendor and local project dependencies, which are used as intended for structural mapping and analysis.
[SAFE]: The skill processes untrusted user-provided text as its primary function. While this creates an indirect prompt injection surface, it is considered safe because it includes a mandatory human-in-the-loop review process and tags all resulting data with '[导入反推]' to prevent accidental reliance on unverified AI output. 1. Ingestion points: Novel text and file paths provided by the user in Phase 1 of SKILL.md. 2. Boundary markers: Content tags are present in the final output; explicit boundary markers are absent during the intermediate analysis steps. 3. Capability inventory: Spawning subagents (story-explorer) and writing project files (SKILL.md Phase 4). 4. Sanitization: No content filtering is applied to the novel text.

story-import