prompt-optimize

Pass

Audited by Gen Agent Trust Hub on Jun 23, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill's primary function is to execute a user-specified shell command (eval_cmd) in a loop to measure prompt performance. This command is treated as a black-box oracle by the agent.
  • [PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface through its feedback-driven optimization loop.
  • Ingestion points: The agent reads the output of the evaluation command, specifically a feedback field in a JSON object or raw text on the last line, as described in SKILL.md.
  • Boundary markers: There are no boundary markers or instructions to isolate the evaluation feedback from the agent's internal instruction processing.
  • Capability inventory: The agent can execute shell commands (eval_cmd) and perform file system operations (reading/writing the <prompt_file> and creating snapshots in the <sandbox_root>).
  • Sanitization: The instructions do not specify any validation or sanitization of the feedback content before it is used to diagnose prompt weaknesses and apply structural edits.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 23, 2026, 04:43 PM
Security Audit — agent-trust-hub — prompt-optimize