self-review
Pass
Audited by Gen Agent Trust Hub on Nov 10, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill evaluates untrusted user-supplied manuscript files and data artifacts, creating a surface for indirect prompt injection attacks.
- Ingestion points: User manuscripts (markdown), extraction CSVs, and configuration JSONs are read and processed across multiple checking phases (Phase 1, Phase 2.5).
- Boundary markers: No specific delimiters or instructions to ignore embedded commands within the analyzed manuscript text were identified.
- Capability inventory: The agent performs local file reads, shell command execution (find, grep), and subprocess calls to specialized Python scripts. It can also edit the manuscript directly when the --fix flag is enabled.
- Sanitization: The skill relies on regular expression matching to extract data from manuscripts but lacks specific sanitization or filtering for adversarial prompt injection strings.
Audit Metadata