The Agent Skills Directory

[COMMAND_EXECUTION] (LOW): The skill utilizes several local bash scripts (scripts/progress.sh, scripts/01-check-environment.sh, scripts/02-verify-deploy.sh) to manage the wizard's state machine and verify environment readiness. These are integral to the skill's primary function.
[EXTERNAL_DOWNLOADS] (LOW): The skill downloads and executes remote code via npx create-next-app and installs the agentation package from npm. While these are external dependencies from non-whitelisted sources, they are standard tools for the development workflow described.
[PROMPT_INJECTION] (LOW): (Indirect) The skill uses WebFetch in steps/03-gather-idea.md to analyze user-provided reference URLs. This is an indirect injection surface where a malicious site could attempt to influence the agent's code generation.
Ingestion points: steps/03-gather-idea.md (WebFetch tool).
Boundary markers: Absent; the instructions do not specify delimiters for the fetched content.
Capability inventory: Full shell execution, package installation, and automated deployment via Vercel.
Sanitization: None; the agent is instructed to extract patterns and use them directly to guide the build.
[DYNAMIC_EXECUTION] (LOW): In steps/04-scaffold-and-build.md, the agent dynamically generates React components (app/agentation-wrapper.tsx) and modifies existing files. This is consistent with the primary purpose of a code-generation assistant.

cc4d