agent-manager-skill

Fail

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: HIGHEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill requires cloning a repository from GitHub account fractalmind-ai, which is not a recognized trusted vendor.
  • [REMOTE_CODE_EXECUTION]: The skill directs the execution of Python scripts (main.py) directly from the cloned external repository.
  • [COMMAND_EXECUTION]: The skill uses python3 and tmux to run and monitor background agent sessions on the local system.
  • [INDIRECT_PROMPT_INJECTION]: 1. Ingestion points: Data is ingested through the assign command here-doc in SKILL.md. 2. Boundary markers: No delimiters or ignore instructions are present. 3. Capability inventory: The skill manages background sessions and monitors agent output. 4. Sanitization: No sanitization logic is visible in the instructions.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Sep 15, 2026, 01:25 PM
Security Audit — agent-trust-hub — agent-manager-skill