hang-check

Pass

Audited by Gen Agent Trust Hub on Jul 30, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill implements a surface for indirect prompt injection by design. It evaluates the health of subagents by reading their logs and output, which could contain instructions meant to deceive the watchdog.
  • Ingestion points: The skill reads recent lines of execution history via the TaskOutput tool and monitors file modification times in the work area to determine if progress is being made.
  • Boundary markers: No explicit delimiters or instructions are provided to the agent to disregard commands or formatting embedded within the subagent output it reads.
  • Capability inventory: The skill possesses the capability to terminate active processes using TaskStop and to persist across sessions using the CronCreate and CronDelete tools.
  • Sanitization: The skill lacks mechanisms to sanitize or validate the content of the monitored output, increasing the risk that a subagent could influence the watchdog's behavior through crafted logs.
  • [SAFE]: The skill follows operational best practices by requiring manual investigation of suspected hangs and ensuring that scheduled monitoring jobs are removed when tasks are complete, which mitigates the risk of accidental resource exhaustion.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 30, 2026, 04:44 PM