diagnose

Pass

Audited by Gen Agent Trust Hub on Jul 10, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to run various local diagnostic tools and scripts, such as test runners, curl, git bisect, and custom bash scripts. These operations are restricted to the local development environment and are necessary for the skill's purpose.
  • [PROMPT_INJECTION]: The skill captures user-provided input, including error messages, via scripts/hitl-loop.template.sh. This creates a surface for indirect prompt injection if the user provides malicious instructions disguised as error data.
  • Ingestion points: User responses are captured into the ERROR_MSG and ERRORED variables within scripts/hitl-loop.template.sh.
  • Boundary markers: Absent; the script outputs raw captured text for the agent to parse without explicit delimiters or warnings to ignore embedded instructions.
  • Capability inventory: The agent has permissions to execute shell commands, modify source code, and perform network requests.
  • Sanitization: No sanitization or validation of the captured input is performed before it is presented to the agent.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 10, 2026, 06:43 AM
Security Audit — agent-trust-hub — diagnose