oma-scholar
Pass
Audited by Gen Agent Trust Hub on Oct 11, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external research content and YAML sidecars, creating a potential attack surface for indirect prompt injection.\n
- Ingestion points: Research papers, abstracts, and third-party .knows.yaml files are processed during generation, validation, and analysis modes (SKILL.md, resources/execution-protocol.md).\n
- Boundary markers: The skill implements clear boundary markers, specifically in resources/upstream-spec-cache.md, where the agent is instructed to treat cached specification snapshots as reference data only and not to obey any imperative-sounding text within them.\n
- Capability inventory: The skill uses local CLI commands (oma scholar), scholarly API requests (OpenAlex, Semantic Scholar), and local file system operations to write sidecars.\n
- Sanitization: A dedicated scripts/sanitize.py tool is included to clean output artifacts and resolve potential extraction hallucinations.\n- [EXTERNAL_DOWNLOADS]: The skill fetches academic metadata from well-known scholarly platforms including OpenAlex and Semantic Scholar (Allen Institute for AI). These resources are reputable services standard to the scholarly research domain.
Audit Metadata