The Agent Skills Directory

[METADATA_POISONING]: The skill's YAML frontmatter includes an authoritative safety claim ('risk: safe'). Per security analysis protocols, self-asserted safety claims in skill metadata are evaluated as data points and do not influence the final verdict, as they could be used to mislead automated or human review processes.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process user requests for landing page generation, creating a vulnerability surface where malicious instructions embedded in user data could influence the agent's behavior.
Ingestion points: User requests are processed in SKILL.md to determine numerical seeds and design parameters.
Capability inventory: The skill has the capability to generate and suggest the execution of React/JavaScript code involving external libraries (GSAP, Phosphor Icons).
Boundary markers: There are no explicit delimiters or instructions provided to the agent to disregard potentially malicious instructions within the user-provided prompt.
Sanitization: The skill does not define sanitization or validation logic for external input before it is used to generate UI code.
[EXTERNAL_DOWNLOADS]: The skill references well-known external libraries and assets for its primary functionality.
Recommends the use of @gsap/react and @phosphor-icons/react for frontend components.
Directs the agent to fetch placeholder imagery from https://picsum.photos.

gpt-taste