The Agent Skills Directory

[SAFE]: The skill instructions define a legitimate process for novel development and review without any malicious intent or safety bypasses.
[DATA_EXFILTRATION]: No exfiltration patterns were identified. The skill reads project-specific documents such as '正文.md' and '主角档案.md' and writes reports to local directories like '工作区/' or '相关设定/'. There is no access to sensitive system paths or external network transmission.
[COMMAND_EXECUTION]: No shell commands, system utilities, or script execution patterns are present. The skill's functionality is limited to reading text files and producing markdown reports.
[PROMPT_INJECTION]: An indirect prompt injection surface is present due to the ingestion of user-generated content, though the risk is mitigated by the skill's lack of dangerous capabilities.
Ingestion points: Manuscript chapters ('正文.md'), world settings ('世界设定.md'), and character profiles ('主角档案.md').
Boundary markers: Absent; the agent follows structured markdown templates for its output reports.
Capability inventory: Reading and writing local markdown files within the defined novel project structure.
Sanitization: Absent; the agent focuses on identifying narrative and setting inconsistencies.

novel-review