novel-proofreader
Pass
Audited by Gen Agent Trust Hub on Oct 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data in the form of Japanese novel manuscripts and user-defined rule files.
- Ingestion points: The agent reads manuscript files (TXT, Markdown) and optional user-provided configuration files such as character lists or
expression-rules.mdas outlined inSKILL.mdandreferences/proofreading-checks.md. - Boundary markers: The skill does not define explicit delimiter-based boundaries (e.g., XML tags) for the input content, although it follows structured typesetting and proofreading logic that defines how the agent should interpret different sections.
- Capability inventory: The skill utilizes file system tools including
Read,Grep,Write, andEditto analyze, modify, and generate manuscript files and reports. - Sanitization: While no general input sanitization is present,
references/proofreading-checks.md(Section 5) contains a specific requirement to detect "meta-expressions," which includes identifying instructions intended for the AI or remnants of chat history, providing a targeted mitigation against certain injection patterns.
Audit Metadata