The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses uv run to execute local Python scripts (resolve_customization.py, resolve_config.py, memlog.py, brain.py) for session management and configuration resolution.
[COMMAND_EXECUTION]: Instructs the agent to use platform-appropriate commands (open, xdg-open, or start) to launch a locally generated HTML selection page in the user's web browser.
[INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by processing user-supplied topics and ideation content to generate imaginative HTML keepsakes and Markdown intent documents.
Ingestion points: User-provided topic, goal, and brainstorming ideas are ingested in SKILL.md and references/mode-partner.md.
Boundary markers: Uses a structured session log (.memlog.md) managed via memlog.py to maintain state and delimit user input from coach instructions.
Capability inventory: Subprocess execution via uv run for internal scripts, file system writes via subagents, and browser-based rendering of generated HTML.
Sanitization: The brain.py script escapes HTML characters when generating the selector page; however, generated artifacts (HTML keepsakes) are dynamically created by subagents based on the session history.
[EXTERNAL_DOWNLOADS]: The module-manifest.toml file identifies github.com/bmad-code-org/bmad-skills as the official update source for the skill.

bmad-brainstorming