supervisor
Pass
Audited by Gen Agent Trust Hub on Aug 10, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill processes untrusted user input by interpolating a user-provided goal into a foundational brief for a secondary autonomous agent (Root). This creates an indirect prompt injection surface where a malicious user goal could influence the downstream agent's behavior.\n
- Ingestion points: The
<user goal>placeholder inSKILL.mdis directly interpolated into theROOT_BRIEFused to initialize thecodex-rootagent.\n - Boundary markers: Absent. The skill does not specify any delimiters or instructions to the downstream agent to ignore potentially malicious commands embedded within the user-supplied goal.\n
- Capability inventory: The skill instructions specify launching the
codex-rootagent with 'full-access', which significantly increases the potential impact of a successful injection attack.\n - Sanitization: Absent. There is no mention of validation, filtering, or sanitization of the user input before it is passed to the next agent.
Audit Metadata