The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local Python scripts using the uv run command for internal workflow management. Specifically, it calls resolve_customization.py, resolve_config.py, and memlog.py located within the {project-root}/_bmad/scripts/ directory to handle configuration and session persistence.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze external content including user-provided transcripts, memos, decks, and research reports during the 'Discovery' phase.
Ingestion points: External source materials (memos, decks, transcripts, Slack threads) and web research digests provided by subagents.
Boundary markers: Employs relevance-filtered extraction to process inputs, though it lacks explicit syntactic delimiters for third-party content.
Capability inventory: File system writes to the workspace, execution of local management scripts, and utilization of MCP tools for external handoffs.
Sanitization: Content is relevance-filtered by subagents before being presented to the parent agent.
[DATA_EXFILTRATION]: The skill includes functionality for 'External Handoffs,' which routes produced artifacts to organization-controlled destinations like Confluence or Slack. These operations use specific MCP tools (e.g., corp:confluence_upload) and are triggered at the finalization stage based on organization or user configuration.

bmad-product-brief