The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses uv run and subprocess.run to execute local Python scripts (such as resolve_customization.py, resolve_config.py, memlog.py, and roster.py) that are part of the BMad ecosystem. These executions are used to resolve configurations and maintain session state on disk.
Evidence: SKILL.md contains multiple uv run commands targeting scripts in the {project-root}/_bmad/scripts/ directory.
Evidence: scripts/resolve_personas.py uses subprocess.run to call other BMad utility scripts using the current Python interpreter.
[INDIRECT_PROMPT_INJECTION]: The skill functions by ingesting user-provided ideas and existing project materials to facilitate a questioning session. This creates a surface where malicious instructions could be embedded in the data being processed.
Ingestion points: User ideas and goals provided during activation in SKILL.md, project files and materials referenced in SKILL.md under 'The forge', and existing .memlog.md files read during the 'On Activation' sequence.
Boundary markers: The instructions do not explicitly define delimiters or provide 'ignore instructions' warnings when processing external project files.
Capability inventory: The skill can read from the filesystem (globbing and reading memlogs), write session logs and HTML reports, and execute local BMad management scripts through the command line.
Sanitization: No specific sanitization or escaping mechanisms are described for user-provided ideas before they are passed to the memlog.py utility or included in the agent's reasoning context.

bmad-forge-idea