The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill documentation describes patterns for ingesting untrusted data from external sources like API responses, form inputs, and URL parameters.
Ingestion points: External data enters the application context through fetch calls in examples/core.md and examples/advanced-patterns.md, as well as URLSearchParams in examples/transforms.md.
Boundary markers: The instructions in SKILL.md explicitly mandate the use of safeParse to handle invalid inputs gracefully and recommend z.strictObject() to reject unknown keys at the trust boundary.
Capability inventory: The skill contains only documentation and static TypeScript code snippets; no executable capabilities such as shell command execution, subprocess spawning, or file system writes are present.
Sanitization: Robust sanitization and validation are the primary focus of the skill, utilizing Zod's refine, transform, and coerce mechanisms to ensure data integrity before further processing.
[EXTERNAL_DOWNLOADS]: The skill's metadata refers to a configuration schema hosted on the author's official GitHub repository (github.com/agents-inc). Code examples demonstrate standard and expected API communication using fetch to generic internal endpoints, representing standard and safe development patterns.

web-forms-zod-validation