The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses standard shell utilities such as find, grep, and wc, along with language-specific build tools like go, npm, python, and make. These commands are used locally to analyze the repository's structure and verify the generated infrastructure, following common development workflows.
[REMOTE_CODE_EXECUTION]: The skill generates local linter scripts and environment setup scripts which are subsequently executed. It also references well-known official Docker images (e.g., PostgreSQL, Redis) for setting up development environments. All execution is transparent and scoped to the project's local development lifecycle.
[DATA_EXFILTRATION]: Analysis revealed that the skill actively scans for environment variables and potential secrets to ensure they are handled correctly. It explicitly mandates the use of environment variable references (e.g., ${VAR_NAME}) and forbids hardcoding sensitive data. No patterns indicative of data exfiltration or unauthorized network communication were found.
[PROMPT_INJECTION]: The skill's instructions and sub-agent prompts do not contain any attempts to override system behavior, bypass safety filters, or extract sensitive model information.

harness-creator