academic-research-skills-codex
Pass
Audited by Gen Agent Trust Hub on Sep 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data such as research manuscripts, draft notes, and reviewer feedback provided by the user. This establishes a surface for indirect prompt injection where instructions embedded in those documents could attempt to influence the agent's behavior.
- Ingestion points: User-provided attachments including
manuscript.pdf,notes.md, andreviewer_comments.txt(SKILL.md). - Boundary markers: The skill does not define explicit delimiters or instructions to ignore embedded commands within the processed materials.
- Capability inventory: The skill can read local files, perform optional network calls for cross-model review, and generate structured output.
- Sanitization: No specific sanitization or validation of the ingested external content is described.
- [EXTERNAL_DOWNLOADS]: The skill includes instructions to download its core components from a specific GitHub repository (
Imbad0202/academic-research-skills-codex). While this involves fetching external code, it is the intended primary installation method and uses standard version control hosting. - [DATA_EXFILTRATION]: The skill supports optional cross-model verification by sending content to the Anthropic API. This is a functional feature triggered by user configuration rather than a silent exfiltration pattern.
Audit Metadata