The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from the project directory to generate implementation rules.
Ingestion points: In steps/step-01-discover.md, the agent is instructed to read architecture documents (game-architecture.md, architecture.md), engine configuration files, and source code patterns from the {project-root}.
Boundary markers: The skill does not implement explicit boundary markers or 'ignore' instructions when reading these files.
Capability inventory: The skill writes the extracted information to {output_folder}/project-context.md, which is specifically intended to be used as instructions for subsequent AI agents. It also has the capability to execute local scripts.
Sanitization: There is no evidence of sanitization or filtering of the content ingested from the project files before it is written to the new context file.
[COMMAND_EXECUTION]: The skill executes a local Python script to manage its workflow configuration.
Evidence: SKILL.md and steps/step-03-complete.md both contain instructions to run python3 {project-root}/_bmad/scripts/resolve_customization.py. This script is part of the internal toolchain for the project's environment and is used to merge configuration overrides.

gds-generate-project-context