The Agent Skills Directory

[COMMAND_EXECUTION]: The assets/collision-lab/tools/check_collision.py script utilizes subprocess.run to execute make_art.py and invoke the host C compiler (such as clang or gcc) to build validation binaries.
[COMMAND_EXECUTION]: The assets/collision-lab/README.md file contains instructions for building the project using the ewrogers/megadev:latest Docker image.
[DYNAMIC_EXECUTION]: The validation tool assets/collision-lab/tools/check_collision.py performs runtime compilation of the project's C source files and executes the resulting binaries to verify collision math and contracts across different optimization levels.
[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes geometric data from assets/collision-lab/res/layouts.json. The tools/make_art.py script parses this JSON to generate art assets and C header files.
Ingestion points: assets/collision-lab/res/layouts.json consumed by tools/make_art.py.
Boundary markers: None identified.
Capability inventory: File writing in make_art.py; subprocess execution and compilation in check_collision.py.
Sanitization: make_art.py performs basic range checks on wall counts and coordinate bounds.

megadev-collision