The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the game's documentation (GDD, Epics, Game Brief, Narrative) to drive architectural decisions. It lacks explicit boundary markers or sanitization, potentially allowing malicious instructions embedded in these documents to influence the agent's behavior during the design process.
Ingestion points: SKILL.md (persistent facts loading), steps/step-02-context.md (loading GDD, Epics, Brief, and Narrative files).
Boundary markers: Absent in instructions for document reading.
Capability inventory: File writing, command execution (uv run), and platform tools (WebSearch).
Sanitization: Absent.
[DYNAMIC_EXECUTION]: The skill executes a Python script (resolve_customization.py) located within the project's _bmad/scripts/ directory. This script is used to resolve customization settings at skill load time and completion, meaning it executes code from the project root being processed.
Evidence: SKILL.md and steps/step-09-complete.md contain instructions to run uv run {project-root}/_bmad/scripts/resolve_customization.py.
[COMMAND_EXECUTION]: The workflow relies on uv run to execute local project scripts for configuration management and workflow completion logic. While these scripts are part of the vendor's expected project structure (_bmad), they execute within the project environment.

gds-game-architecture