The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses platform-specific shell commands to open generated HTML artifacts for user review. Specifically, it invokes open "PATH" on macOS, xdg-open on Linux, and start "" on Windows as documented in assets/color-themes.md, references/creative-tools.md, and references/validate.md. Additionally, the skill executes local Python scripts using uv run for workspace initialization and configuration resolution (e.g., resolve_customization.py, memlog.py) as seen in SKILL.md.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process potentially untrusted data to generate UX specifications.
Ingestion points: External source files, user-supplied visuals, sketches, and brand decks located in the imports/ and {planning_artifacts}/ directories as specified in SKILL.md.
Boundary markers: The skill utilizes specialized subagents for data extraction and a "rubric walker" validation process to structure input, though these do not provide complete sanitization against adversarial content.
Capability inventory: The agent has the ability to write to the filesystem (DESIGN.md, EXPERIENCE.md, .working/ directory) and execute shell commands for previewing files.
Sanitization: There is no evidence of explicit escaping or filtering of content extracted from user-supplied documents before it is interpolated into the generated markdown spines.

bmad-ux