The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill involves the use of external game engine frameworks such as Bevy, Flecs, and Unity DOTS. The instructions advise pinning specific versions and verifying packages against schemas to mitigate supply chain risks (ASI04).
[COMMAND_EXECUTION]: The skill is configured to use the execute_command tool alongside build and test runners. This access is leveraged for executing performance benchmarks and verifying deterministic simulation outputs.
[DYNAMIC_EXECUTION]: The skill addresses risks associated with dynamic execution by prohibiting the composition of schedules and queries from dynamic strings or external configurations without validation (ASI05).
[INDIRECT_PROMPT_INJECTION]: The skill identifies a potential attack surface where external component data or simulation snapshots could be used to poison the agent's context.
Ingestion points: The agent reads data from files via read_file and search_code during architecture implementation.
Boundary markers: No specific delimiters or boundary warnings are present in the skill instructions.
Capability inventory: The skill allows write_file, execute_command, run_build, and create_file across all implementation steps.
Sanitization: The skill explicitly includes security guardrails to validate component payloads against component schemas and treat all deserialized state as untrusted (ASI06).

architect-ecs-systems