The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted project artifacts such as user stories (story_file), PRDs, and technical specifications. These inputs are used to generate executable test code and Maestro flows, creating an attack surface where malicious instructions in project documentation could lead to the generation of harmful code. Evidence found in SKILL.md and steps-c/step-01-preflight-and-context.md.
[DYNAMIC_EXECUTION]: The run-and-heal.md resource describes a loop where the agent executes the generated test files using the local project runner. This involves the dynamic execution of code created by the agent during the current session, managed by the auto_heal_failures and auto_validate settings.
[COMMAND_EXECUTION]: The skill executes local Python scripts using uv run at paths like {project-root}/_bmad/scripts/resolve_customization.py. This relies on the existence of specific scripts within the project's own directory structure, as seen in SKILL.md Step 1.
[EXTERNAL_DOWNLOADS]: The skill invokes npx to download and execute packages, including bmad-method-test-architecture-enterprise and other skills from the bmad-code-org repository. These actions reference resources belonging to the skill's author (bmad-code-org).

bmad-tea-automate