research-ops-skills
Pass
Audited by Gen Agent Trust Hub on Oct 24, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [COMMAND_EXECUTION]: The skill requires the agent to run local Python scripts, specifically onboard.py for interactive setup and ar_evaluator.py for automated evaluation tasks.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from multiple sources such as interview transcripts, survey exports, and protocol drafts, which could contain hidden instructions.
- Ingestion points: SKILL.md references processing of protocol drafts, program ledgers, survey exports, and interview transcripts.
- Boundary markers: No explicit boundary markers or instruction-ignoring prompts are defined for the intake of external data.
- Capability inventory: The skill has the ability to execute Python scripts via shell commands and perform file operations in the workspace.
- Sanitization: There is no evidence of sanitization or filtering of the content within the research artifacts.
- [DATA_EXFILTRATION]: The skill accesses and processes sensitive corporate data including R&D budgets, clinical trial protocols, and market sizing models. It also stores user configuration in the ~/.config/research-ops/ directory.
Audit Metadata