manuscript-provenance
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from user-provided manuscript sources and codebases.
- Ingestion points: Processes multiple file types including LaTeX sources (.tex), scripts (.py, .sh), and configuration files (.json, .yaml, .toml, Makefile) to build registries and trace provenance (Phase 1a, 1b).
- Boundary markers: There are no specific instructions or delimiters defined to prevent the agent from obeying instructions that might be embedded within the manuscript text or code comments.
- Capability inventory: The skill possesses the capability to read any file within the provided project directory and write report files back to the local environment (Phase 6).
- Sanitization: The instructions do not specify any validation or sanitization of the data extracted from the ingested files before it is used for report generation or manifest creation.
Audit Metadata