mpm-circuit-breaker-enforcement

Pass

Audited by Gen Agent Trust Hub on Mar 30, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill defines a set of restrictive guardrails designed to prevent the primary agent from overstepping its intended role. It explicitly blocks high-risk actions like direct source code modification (Circuit Breaker #1), excessive file investigation (Circuit Breaker #2), and direct execution of verification commands like curl or lsof (Circuit Breaker #7).
  • [SAFE]: No malicious patterns such as prompt injection, data exfiltration, or obfuscation were detected. The use of instructional language like 'BLOCK' and 'REQUIRE' is applied in a benign context to enforce process integrity rather than to bypass safety filters.
  • [SAFE]: The skill includes protection against social engineering by flagging attempts to delegate technical tasks back to the user (Circuit Breaker #9), ensuring the agent remains within its operational boundaries.
  • [SAFE]: The skill enforces a secure software development lifecycle (SDLC) by requiring git tracking (Circuit Breaker #4) and mandatory QA verification (Circuit Breaker #8) before tasks are marked as complete.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 30, 2026, 02:30 PM
Security Audit — agent-trust-hub — mpm-circuit-breaker-enforcement