The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill implements a lead capture system that ingests untrusted user input via web forms. This data is intended for storage and subsequent notification or AI-assisted responses.
Ingestion points: User data enters via the submitLead server action and is validated against the leadInputSchema in SKILL.md.
Boundary markers: While the skill describes the data flow, it does not include the specific prompt templates where the data is used, which are reportedly managed by related skills.
Capability inventory: The skill is capable of database persistence (createLead) and triggering external tasks (trigger.dev).
Sanitization: The implementation uses the zod library to strictly enforce types and length constraints on all user-submitted fields.
[SAFE]: The skill demonstrates security best practices, such as using timingSafeEqual for secret comparison and providing clear warnings about potential IP spoofing in various hosting environments.

launch-lead-capture