The Agent Skills Directory

Unverifiable Dependencies & Remote Code Execution (LOW): The skill provides instructions to install packages from the @tanstack organization (e.g., @tanstack/react-form, @tanstack/react-devtools). While reputable in the JavaScript community, this organization is not on the strictly defined trusted sources list. The use of floating versions (@latest) is noted as a supply chain consideration. The severity is reduced to LOW as these dependencies are central to the skill's primary educational purpose.
Indirect Prompt Injection (LOW): The skill documentation describes patterns for creating forms to ingest untrusted user input, creating a potential attack surface. 1. Ingestion points: Input components illustrated in rules/ui-libraries.md and rules/arr-array-fields.md. 2. Boundary markers: No explicit delimiters or instructions to ignore embedded commands are suggested in the code examples. 3. Capability inventory: Use of fetch in onSubmit handlers and server-side logic patterns in rules/ssr-meta-frameworks.md. 4. Sanitization: The skill promotes the use of schema validation (e.g., Zod) in rules/val-schema-validation.md, which validates data structure but does not mitigate malicious instruction intent within natural language fields.

tanstack-form