The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell commands (e.g., find, ls, grep, wc) to detect project state, tech stack, and code architecture during the analysis phases.
[COMMAND_EXECUTION]: The verification phase (Phase 5) runs project-specific build and test commands such as go build ./..., npm run build, and make lint-arch to ensure the generated infrastructure is valid.
[EXTERNAL_DOWNLOADS]: The skill creates CI configurations (e.g., .github/workflows/ci.yml) and verification steps that utilize standard package managers (Go modules, npm, pip) to install dependencies from official registries.
[DYNAMIC_EXECUTION]: The skill generates architectural linters and environment setup scripts (e.g., scripts/lint-deps.go, harness/scripts/setup-env.sh) from internal templates and then executes them as part of the harness verification process.
[SAFE]: The skill incorporates explicit safety warnings and best practices, such as instructing subagents never to hardcode secrets and to use environment variables (${VAR_NAME}) for sensitive configuration.
[SAFE]: The skill follows a transparent workflow with parallel subagents for specific tasks (analysis, documentation, config), providing a structured approach to infrastructure generation without hidden or malicious side effects.

harness-creator