The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes repository data, such as manifests and integration code, which serves as an ingestion point for untrusted content.
Ingestion points: Repository files accessed via the Read, Glob, and Grep tools.
Boundary markers: The instructions do not specify explicit data delimiters, but they mandate a "read-only discovery" phase and treat repository content as evidence rather than instructions.
Capability inventory: The skill possesses Write and Edit tools for file modification and WebFetch for network access.
Sanitization: No specific content sanitization is mentioned, but the skill includes strong procedural guardrails, requiring explicit user authorization for all mutations and fresh approval for production publication.
[EXTERNAL_DOWNLOADS]: The skill retrieves technical documentation and status information from official Webflow domains.
Evidence: Fetches data from developers.webflow.com and status.webflow.com to verify current API contracts and service health.

webflow-incident-runbook