The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses uv run to execute local utility scripts such as memlog.py, resolve_customization.py, and resolve_config.py located within the project's _bmad/scripts/ directory. These are used for maintenance tasks like logging decisions and resolving configurations.
[COMMAND_EXECUTION]: The validation workflow includes an instruction to open a generated HTML report in the system's default browser using platform-specific commands like open, xdg-open, or start.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and summarize large amounts of external, untrusted content including web research, customer transcripts, and competitor documentation. This data ingestion creates a surface for indirect prompt injection attacks.
Ingestion points: User-provided verbal context, product briefs, external research, and transcripts (Discovery section).
Boundary markers: Uses subagents for extraction to isolate processing, but lacks explicit sanitization of the extracted text before interpolation.
Capability inventory: File system writes, local script execution via uv run, browser invocation, and network research.
Sanitization: No specific evidence of escaping or sanitizing data interpolated into internal commands or prompts.
[EXTERNAL_DOWNLOADS]: The skill mentions external dependencies and setup steps using npx and GitHub repositories. These resources are associated with the skill's author and represent normal functional requirements.

bmad-prd