k6-test-maintenance
Pass
Audited by Gen Agent Trust Hub on Aug 4, 2026
Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill analyzes external k6 test scripts which represent an untrusted data source. This creates a surface for indirect prompt injection where malicious instructions embedded in script comments or code could attempt to influence the agent's behavior.
- Ingestion points: Local file system scripts and cloud-hosted scripts fetched via k6-manage.
- Boundary markers: Uses behavior-aware change control and human-in-the-loop confirmation for non-syntactic changes, but lacks explicit delimiters to isolate untrusted script content.
- Capability inventory: Has access to gcx for cloud API access, k6-manage for script modification, and the k6 CLI for script execution.
- Sanitization: Validates script syntax using k6 inspect and validate_script before proposing or applying changes.
- [EXTERNAL_DOWNLOADS]: Fetches k6 utility functions from official repositories and accesses documentation from Grafana domains.
- Evidence: Imports from https://jslib.k6.io/k6-utils/1.5.0/index.js and documentation lookups at https://grafana.com/docs/k6/latest/.
- [REMOTE_CODE_EXECUTION]: Executes k6 scripts both locally and in Grafana Cloud as part of the standard post-edit verification process.
- Evidence: Commands like k6 run and k6 cloud run are used to ensure script integrity.
- [COMMAND_EXECUTION]: Utilizes the k6 and gcx CLIs to manage test environments, inspect code, and retrieve authentication tokens.
- Evidence: Uses gcx --context config view and k6 inspect for parse checking.
Audit Metadata