shared
Pass
Audited by Gen Agent Trust Hub on Feb 19, 2026
Risk Level: SAFE
Full Analysis
- PROMPT_INJECTION (SAFE): While the skill defines protocols for inter-agent communication and includes templates for sub-agent role-play (e.g., 'You are judge-1'), these are standard coordination patterns. The use of delimited blocks like
<PACKET>in templates provides structural separation, reducing the risk of accidental instruction confusion. - COMMAND_EXECUTION (SAFE): The skill contains a validation script (
scripts/validate.sh) and provides examples for using thecodex execandBashtools. These are restricted to standard development tasks and do not show signs of malicious command injection or privilege escalation. - EXTERNAL_DOWNLOADS (SAFE): The documentation refers to a Go package (
github.com/boshu2/agentops) for parsing protocol messages. This reference is informational only and no automated installation or execution of untrusted external code is performed by the skill.
Audit Metadata