The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a Python script via uv run to resolve customization settings (resolve_customization.py). This script is located within the project's own infrastructure ({project-root}/_bmad/scripts/) and is used for standard configuration merging (base, team, and user overrides).
[COMMAND_EXECUTION]: A similar execution occurs at the end of the workflow (Step 6) to handle on_complete logic. Both instances use path variables that resolve to the project's local workspace, representing standard developer tooling behavior.
[DATA_EXPOSURE]: The skill reads various project artifacts (GDD, Epics, Architecture, UX Design) to assess the impact of changes. This is consistent with its primary purpose of course correction and change management. No sensitive system files or credentials (e.g., .ssh, .aws, .env) are accessed.
[REMOTE_CODE_EXECUTION]: There are no remote code execution patterns. All scripts and tools referenced are local to the project root or the skill root.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data in the form of project artifacts (GDD, Narrative, etc.) and user-provided issue descriptions. However, it implements a highly structured, interactive checklist and requires explicit user approval before finalizing or routing any changes, which serves as a human-in-the-loop sanitization layer.

gds-correct-course