The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local Python scripts (resolve_customization.py) found in the {project-root}/_bmad/scripts/ directory. These executions occur during the activation phase in SKILL.md and upon workflow completion in step-12-complete.md to resolve customization blocks and execute final actions.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface due to its deep integration with external project data.
Ingestion points: Multiple project locations are scanned in step-01-init.md, including {planning_artifacts}, {project_knowledge}, and {implementation_artifacts}, as well as customize.toml for persistent facts.
Boundary markers: Absent; content from discovered files is loaded directly into the agent's memory for processing.
Capability inventory: The ingested data significantly influences the generation of success criteria, user journeys, and functional requirements throughout the workflow.
Sanitization: No evidence of sanitization or filtering of the content from the project markdown or CSV files is present.
[DATA_EXPOSURE]: The skill performs extensive file system discovery and content loading from various project subdirectories (e.g., investigations, research, and project-context files) to populate the context for the generated PRD.

bmad-create-prd