performance-test-k6
Pass
Audited by Gen Agent Trust Hub on Oct 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONPRIVILEGE_ESCALATION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests user-provided materials like SLA documents and traffic logs to generate executable k6 scripts, creating a surface for potential prompt injection through external data.
- Ingestion points: Performance profiles and endpoint lists provided by the user (e.g.,
evals/fixtures/load-profile.mdandprompts/performance-test-k6.md). - Boundary markers: Instructions in
prompts/performance-test-k6.mdrequire the agent to label assumptions and avoid inventing data. - Capability inventory: The skill generates k6 scripts that perform network operations (
http.get,http.post) and uses local scripts to process test results. - Sanitization: Prompt guidelines mandate using
__ENVfor secrets and placeholders for tokens to prevent credential hardcoding. - [COMMAND_EXECUTION]: The evaluation framework in
evals/graders/review-offline-arrival-artifacts.pyusessubprocess.runto callnode --checkfor validating the syntax of generated k6 scripts. - [PRIVILEGE_ESCALATION]: Installation instructions in
README.mdfor Windows suggest using-ExecutionPolicy Bypassfor PowerShell, which is a standard method for installers but involves bypassing default execution restrictions.
Audit Metadata