The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses uv run to execute internal configuration and customization scripts (e.g., resolve_customization.py, resolve_config.py) located within the project's _bmad/scripts/ directory. This is a standard architectural pattern for the BMad framework to resolve environment-specific settings.
[REMOTE_CODE_EXECUTION]: The instructions direct the user to run npx commands for installing related components or migrating legacy customizations (e.g., npx --package=bmad-method-test-architecture-enterprise@3 tea-migrate-customizations). These resources originate from the same vendor (bmad-code-org) and are part of the intended setup process.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data including test files, story documents, acceptance criteria, and pull request diffs. It implements safety boundaries by explicitly instructing the agent that such context is read-only and should never be used to waive rubric violations or influence the deterministic scoring logic.
[DATA_EXPOSURE_&_EXFILTRATION]: While the skill reads sensitive project artifacts like configuration files and test code, it does so for analysis purposes. It utilizes local CLI tools like playwright-cli to capture evidence (traces, screenshots) which are stored in local project-relative directories ({test_artifacts}/test-review/). No unauthorized external network transmission of this data was detected.

bmad-tea-test-review