The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is designed to identify and instruct the execution of shell commands for installation, building, and verification (e.g., install, buildOrExport, start, verify) defined within a generated BUILD_BRIEF.md. These commands are determined dynamically based on the project's design and runtime environment.
[INDIRECT_PROMPT_INJECTION]: The skill processes content from untrusted external documents like GAME_DESIGN.md, ART_DIRECTION.md, and PRODUCT_BRIEF.md to determine the build toolchain and execution logic. A malicious design document could influence the generated shell commands.
Ingestion points: GAME_DESIGN.md, ART_DIRECTION.md, PRODUCT_BRIEF.md (specified in SKILL.md and references/build-brief-contract.md).
Boundary markers: None identified in the provided contracts to isolate instructions within these documents.
Capability inventory: Execution of shell commands through the commands section in build-brief-contract.md.
Sanitization: The skill instructions do not specify validation or sanitization of the commands extracted or derived from the design documents.
[DYNAMIC_EXECUTION]: The skill involves runtime generation of execution paths and build scripts. It uses a "signature command" pattern to validate inputs, but the ultimate output is the execution of a candidate build within the target runtime environment.

game-build