cortex-skills-loop

Pass

Audited by Gen Agent Trust Hub on Mar 29, 2026

Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
  • [SAFE]: The skill serves as an instructional guide for the cortex CLI's internal learning and recommendation engine.
  • [COMMAND_EXECUTION]: The skill utilizes several local commands (e.g., cortex skills recommend, cortex status, cortex skills rate) to manage the agent's workflow. These commands are limited to the specific functionality of the developer tool and do not perform unauthorized file access, network exfiltration, or administrative changes.
  • [PROMPT_INJECTION]: The skill monitors the working environment for 'signals' like file patterns and git diff to suggest relevant tools, which constitutes an indirect prompt injection surface.
  • Ingestion points: Monitors git diff and file system patterns to trigger recommendations (SKILL.md).
  • Boundary markers: No explicit markers or warnings are defined for the agent regarding the processing of git output.
  • Capability inventory: Running various cortex CLI subcommands (SKILL.md).
  • Sanitization: None identified; the skill relies on the internal implementation of the cortex CLI to safely handle system data.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 29, 2026, 05:13 PM
Security Audit — agent-trust-hub — cortex-skills-loop