The Agent Skills Directory

[COMMAND_EXECUTION]: The skill makes extensive use of the Bash tool across several specialized sub-agents (Architect, Backend, Frontend, QA, DevOps). This is used for project initialization (via scripts/init-project.sh), running unit/integration/E2E tests, and deploying the resulting application (e.g., npm run dev, docker-compose up). These operations are consistent with the skill's primary purpose as a coding automation tool.
[EXTERNAL_DOWNLOADS]: The PM and Architect agents utilize WebSearch and WebFetch tools to research technology trends, perform competitive analysis, and gather documentation. Additionally, the DevOps and QA agents use package managers like npm, pnpm, pip, and cargo to install necessary dependencies for the project being developed. These downloads are directed at standard registries and requested technology documentation.
[REMOTE_CODE_EXECUTION]: While the sub-agents execute code as part of the development and testing process, the execution is limited to the project's own environment and dependencies. The skill includes a testing gate (references/quality-gate.md) that requires passing automated tests before proceeding to deployment, which serves as a safety mechanism.
[PROMPT_INJECTION]: There is a potential for indirect prompt injection as the skill ingests untrusted data from the web (via WebFetch) or from existing codebases. The sub-agents are instructed to follow strict roles and templates, which provides a layer of structural defense, but the risk remains inherent to any agent processing external content.

boss