The Agent Skills Directory

[COMMAND_EXECUTION]: The skill invokes uv run to execute {project-root}/_bmad/scripts/resolve_customization.py during activation and after completion. These scripts resolve workflow settings and activation steps, which is standard for the author's framework but involves shell interaction with local scripts.
[INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted data from configuration files and user context which could influence its behavior.
Ingestion points: Data is loaded from SKILL.md referenced files, {project-root}/_bmad/custom/ TOML files, and user-supplied context data files.
Boundary markers: No explicit delimiters or safety instructions are used to separate ingested content from system-level instructions to prevent the agent from obeying instructions hidden in data.
Capability inventory: The agent can execute shell commands through uv run based on values resolved from these configuration files, specifically through activation_steps and the on_complete key.
Sanitization: The skill does not perform validation or sanitization on the configuration values before they are used to drive the workflow logic.

bmad-cis-storytelling