The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill downloads dependencies and project boilerplates from official registries such as npm and PyPI. It relies on standard framework initializers like npx (for create-next-app and nuxi), expo-cli, and flutter-cli to bootstrap projects.
[COMMAND_EXECUTION]: Shell commands are executed via Bash to manage project lifecycles, including installing packages, running database migrations, and configuring local development environments across multiple tech stack templates.
[PROMPT_INJECTION]: The skill processes untrusted natural language requests to determine application scope and architecture, creating a potential surface for indirect prompt injection.
Ingestion points: Natural language user requests define the project goals in SKILL.md.
Boundary markers: The orchestration logic enforces a mandatory PLAN.md creation and verification step in agent-coordination.md to structure the agent's work before code generation.
Capability inventory: The skill possesses extensive capabilities including Bash access, file Write/Edit permissions, and the ability to coordinate specialized sub-agents.
Sanitization: Security relies on the structured multi-agent workflow and verification checkpoints rather than explicit programmatic input sanitization.

app-builder