The Agent Skills Directory

[COMMAND_EXECUTION]: The maintenance script evals/build.py utilizes the subprocess.run function to automate type-checking via the TypeScript compiler (tsc) and execute the local verification tool scripts/check.mjs. These executions are performed using list-based arguments without a shell, targeting only internal project files to ensure consistency between the source evaluation definitions and the generated benchmark files.
[EXTERNAL_DOWNLOADS]: The example/package.json file lists standard development dependencies including zod, vitest, and typescript. All libraries are well-known, widely used in the industry, and sourced from official package registries. No suspicious or unverified third-party scripts are downloaded or executed.
[PROMPT_INJECTION]: The SKILL.md instructions are strictly focused on architectural patterns and coding standards for Domain-Driven Design. There are no attempts to bypass AI safety guardrails, override system prompts, or induce restricted behaviors.
[DATA_EXFILTRATION]: There is no evidence of unauthorized data handling. The skill does not access sensitive local file paths (such as SSH keys or cloud credentials) and does not perform network operations to untrusted domains. References to external URLs are limited to schema definitions and example pull request templates.
[SAFE]: The skill follows secure coding practices, utilizing readonly modifiers and Zod's validation capabilities to ensure data integrity. Persistence mechanisms and privilege escalation patterns are absent from the implementation.

implementing-ddd-tactical-patterns-with-zod