drift-testing
Pass
Audited by Gen Agent Trust Hub on Sep 20, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests external data from OpenAPI specifications and Drift datasets to generate test targets and parameters. An adversary providing a malicious specification could potentially influence the agent's behavior during the test generation process.\n
- Ingestion points: OpenAPI spec files (YAML/JSON) parsed by
extract_endpoints.pyandcheck_coverage.py, and test/dataset files processed by the Drift CLI.\n - Boundary markers: The skill does not implement specific boundary markers or instructions to isolate the content of external specifications from the agent's logic.\n
- Capability inventory: The skill performs network operations via the
driftcommand and Luahttp()function, executes subprocesses usinguvandnpx, and performs local file operations.\n - Sanitization: While
yaml.safe_loadis used for parsing, there is no explicit validation or sanitization of the values extracted from the OpenAPI spec before they are used to generate test configurations.\n- [EXTERNAL_DOWNLOADS]: The skill facilitates the acquisition and installation of testing tools from external sources.\n - Evidence: Scripts and documentation provide instructions for downloading the Drift binary from
download.pactflow.ioand installing the Prism mock server from the NPM registry. These sources are recognized as official vendor or well-known service repositories.\n- [DYNAMIC_EXECUTION]: The skill utilizes an embedded Lua environment to allow users to define custom logic for test lifecycle events.\n - Evidence:
references/lua-api.mddetails how to write Lua scripts for hooks likeoperation:startedandhttp:request. These scripts are generated by the agent based on user needs and are executed by the Drift CLI during test runs.\n- [COMMAND_EXECUTION]: The skill relies on executing various system commands and local scripts to automate the testing workflow.\n - Evidence: The skill uses
drift verify,npm install,npx, anduv runto execute the core testing logic and its supporting Python/Shell/PowerShell scripts.
Audit Metadata