load-test-builder
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCREDENTIALS_UNSAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill defines workflows that ingest data from external files (results.json) produced by load testing tools. This creates an attack surface where data from a tested system could influence downstream reporting logic.
- Ingestion points: The analyze-results.js script and GitHub Action workflows parse the results file.
- Boundary markers: No explicit delimiters or warnings are used to isolate potentially untrusted metric data.
- Capability inventory: The skill uses JSON.parse and formatting tools to display results in consoles or PR comments.
- Sanitization: There is no evidence of sanitization for the values extracted from the JSON results.
- [CREDENTIALS_UNSAFE]: The skill includes hardcoded placeholder credentials in its example scripts for local testing scenarios.
- Evidence: artillery/processor.js uses a static password 'testpassword123' for simulated users.
- Evidence: .github/workflows/load-tests.yml contains a local database connection string 'postgresql://test@localhost/test'.
- [COMMAND_EXECUTION]: The skill instructs the agent to perform installation and execution tasks via shell commands.
- Evidence: Instructions cover brew, docker, npm, and apt-get usage.
- [EXTERNAL_DOWNLOADS]: The skill fetches resources from trusted sources to set up the testing environment.
- Evidence: Fetches k6 from the official Grafana repository at dl.k6.io.
- Evidence: Installs Artillery and Autocannon from the standard NPM registry.
Audit Metadata