shatter

Pass

Audited by Gen Agent Trust Hub on Apr 27, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTIONNO_CODE
Full Analysis
  • [COMMAND_EXECUTION]: The skill's procedure instructs the agent to execute potentially destructive system operations, including killing nodes, dropping network packets, and filling disk space, as part of its fault injection methodology.
  • [PROMPT_INJECTION]: The skill utilizes a role-play prompt ('You are the attacker against your own system') to orient the agent's persona toward adversarial testing and brittleness discovery.
  • [NO_CODE]: The skill consists entirely of instructional markdown and YAML metadata; it does not include any scripts, binaries, or other executable components that could run autonomously.
  • [SAFE]: The instructions include mandatory safety guardrails, requiring the agent to restate no-touch boundaries, define a blast radius, confirm authorization, and establish a rollback plan before performing any tests.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 27, 2026, 09:14 AM
Security Audit — agent-trust-hub — shatter