chaos-experiment
Pass
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied data to define and execute chaos experiments, creating a surface for indirect prompt injection.
- Ingestion points: The
<experiment definition>placeholder inSKILL.md(Step 4) accepts untrusted input that is passed directly to theharness_createtool. - Boundary markers: There are no explicit delimiters or instructions to the agent to ignore potentially malicious embedded instructions within the experiment body.
- Capability inventory: The skill has the capability to disrupt infrastructure via
harness_execute(pod deletion, node draining, EC2 stopping) and perform network-related faults. - Sanitization: The instructions do not define validation or sanitization steps for the experiment definition before it is processed by the MCP tools.
Audit Metadata