The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses uv run to execute Python scripts, such as resolve_customization.py and resolve_config.py, located within the project's {project-root}/_bmad/scripts/ directory. These scripts are invoked to manage environment setup and merge configuration files.\n- [DYNAMIC_EXECUTION]: The skill supports a customization system through customize.toml that includes activation_steps and an on_complete hook. The agent is explicitly instructed to execute the content of these fields as terminal instructions, enabling the execution of arbitrary logic or shell commands defined in external project files.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the local file system and the web, creating a vulnerability surface for indirect prompt injection.\n
Ingestion points: The Artifact Analyzer subagent scans project-level directories ({planning_artifacts}, {project_knowledge}) and user-specified file paths to gather context for the PRFAQ.\n
Boundary markers: The instructions do not define specific delimiters or security guardrails to prevent the agent from following malicious instructions embedded within the scanned project documents.\n
Capability inventory: The agent is empowered to execute shell commands via uv run and perform file system writes.\n
Sanitization: There is no evidence of sanitization or validation performed on the external data before it is interpolated into the agent's reasoning process or the final output documents.

bmad-prfaq