skills/openai/plugins/pride-skill/Gen Agent Trust Hub

pride-skill

Pass

Audited by Gen Agent Trust Hub on Sep 20, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATION
Full Analysis
  • Flexible Network Operations: The script scripts/rest_request.py accepts a base_url and path from input, allowing it to perform GET or POST requests to arbitrary URLs. While the skill is intended for the PRIDE Archive (ebi.ac.uk), the underlying code does not restrict the domain, which could be used to reach internal or unintended external services.
  • Local File Writing: The skill supports a save_raw feature that writes API responses to the local filesystem. By default, it writes to /tmp, but the raw_output_path parameter allows specifying other locations. This could potentially be used to overwrite files if the agent has sufficient permissions.
  • Command Execution Surface: The skill is invoked via a shell pipe (echo ... | python scripts/rest_request.py). If the input JSON is dynamically constructed from untrusted data without proper sanitization, it could theoretically lead to command injection, although the current implementation uses standard JSON parsing.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 20, 2026, 12:20 PM
Security Audit — agent-trust-hub — pride-skill