mirrord-chaos
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill provides example shell scripts and commands for local and CI environments to manage mirrord sessions and chaos rules.
- [INDIRECT_PROMPT_INJECTION]: The skill processes external inputs such as JSON rule files and CLI outputs from
mirrord chaos. - Ingestion points:
SKILL.md(CI mode script),mirrord chaos listoutput, and user-provided rule files. - Boundary markers: The skill explicitly instructs the agent in the "Security Boundaries" section to treat command output and rule files as untrusted data and not to execute shell commands derived from their values.
- Capability inventory: Use of
mirrordCLI tools (exec,chaos,session),npm, and shell utilities likeawkin referenced CI examples. - Sanitization: Instructions recommend using JSON format for data handling and specific IDs for rule management to minimize parsing errors.
Audit Metadata