The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to run a local Python script (resolve_customization.py) from the {project-root}/_bmad/scripts/ directory during both the activation (SKILL.md) and completion (step-05-present.md, step-oneshot.md) phases.
[COMMAND_EXECUTION]: Uses the code CLI tool to open the project root and generated specification files in the user's VS Code editor for review (step-05-present.md, step-oneshot.md).
[COMMAND_EXECUTION]: Executes several Git commands, including git status, git rev-parse, and git commit, to manage the local repository state during the workflow (step-01-clarify-and-route.md, step-05-present.md).
[PROMPT_INJECTION]: The skill ingests untrusted project documentation to guide code generation, creating a surface for indirect prompt injection.
Ingestion points: User intent, project-context.md (SKILL.md), and planning artifacts such as PRDs, architecture guides, and epics (step-01-clarify-and-route.md).
Boundary markers: Utilizes <frozen-after-approval> tags in specification templates to separate human-validated intent from agent-derived tasks (spec-template.md, step-02-plan.md).
Capability inventory: Performs file system writes, executes shell commands (Git, Python, VS Code), and delegates work to sub-agents (step-03-implement.md, step-04-review.md).
Sanitization: No explicit sanitization or filtering of ingested document content is described.
[PROMPT_INJECTION]: Dynamically executes instructions defined in the activation_steps_prepend, activation_steps_append, and on_complete configuration fields (SKILL.md, step-05-present.md).

bmad-quick-dev