The Agent Skills Directory

[Indirect Prompt Injection] (LOW): The skill creates an attack surface by interpolating user-provided parameters directly into TypeScript and React component templates.
Ingestion points: User parameters defined in Step 1 (Widget ID, Title, Content, Sponsor URL) within SKILL.md.
Boundary markers: Absent; user-provided strings are placed directly into template literals and JSX attributes.
Capability inventory: The agent is instructed to write .ts and .tsx files to the filesystem (src/lib/plugins/widgets/) and suggests local execution via npm run dev in SKILL.md.
Sanitization: Absent; there is no validation or escaping of the user inputs before they are embedded into the generated source code.
[Dynamic Execution] (LOW): The skill performs script generation from templates. While intended for developer productivity, the assembly of executable files from unverified user strings is a form of dynamic code generation.
[Command Execution] (SAFE): The suggested commands (npx tsc, npm run dev) are standard development workflows and appropriate for the skill's stated purpose of plugin generation.

widget-generator