doc-coauthoring
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFE
Full Analysis
- [Prompt Injection] (SAFE): The instructions focus entirely on document creation workflows and do not attempt to override the underlying AI safety protocols or system instructions.
- [Data Exposure & Exfiltration] (SAFE): There are no attempts to access sensitive local files (like SSH keys or environment variables) or send data to unauthorized external endpoints.
- [Unverifiable Dependencies & Remote Code Execution] (SAFE): The skill does not perform any package installations or execute remote scripts.
- [Indirect Prompt Injection] (LOW):
- Ingestion points: The skill reads external reference materials provided by the user and web search results in Stage 1.
- Boundary markers: Absent. There are no specific delimiters defined to isolate untrusted content from the internal workflow logic.
- Capability inventory: Limited to file reading and web searching. No high-risk capabilities like system-level file writing or arbitrary code execution are present.
- Sanitization: Not explicitly implemented, though the skill focuses on stylistic analysis rather than instruction following from the input data.
- [Obfuscation] (SAFE): The file is written in clear Markdown and standard text with no encoded or hidden characters.
Audit Metadata