research-ops-skills

Pass

Audited by Gen Agent Trust Hub on Jun 13, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands to run local Python scripts, specifically onboard.py and ar_evaluator.py, located within the skill's directory structure. This allows the skill to perform interactive configuration and automated evaluation within the user's environment.
  • [PROMPT_INJECTION]: The skill is designed to ingest and process a wide range of external data, including clinical protocols, financial ledgers, market models, and interview transcripts (in JSON, CSV, and Markdown formats). This creates a vulnerability surface for indirect prompt injection.
  • Ingestion points: The skill actively searches for and reads workspace artifacts such as protocol.json, program-budget.json, tam-model.json, and interview-guide.md to determine routing and provide context to sub-skills.
  • Boundary markers: There are no explicit instructions for the agent to use boundary markers or to ignore embedded instructions within the processed research artifacts.
  • Capability inventory: The skill possesses the capability to execute Python scripts and write configuration data to the local file system (e.g., ~/.config/research-ops/).
  • Sanitization: No sanitization, validation, or structural checking of the ingested research data is described in the skill's instructions.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 13, 2026, 12:17 PM
Security Audit — agent-trust-hub — research-ops-skills