The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill describes functionality that ingests untrusted data, creating a surface for indirect prompt injection.
Ingestion points: Local media files (videos, images, transcripts) and JSON-formatted data sent to visual regression testing endpoints like /@ef-compare-snapshot (SKILL.md).
Boundary markers: No explicit instruction delimiters or boundary markers are defined in the provided setup code.
Capability inventory: The resulting development server can perform network operations (listening on ports 3001/3099) and file system writes (writing cached assets to ./cache and snapshots to {root}/test/__snapshots__).
Sanitization: The provided configuration snippets do not specify sanitization or validation logic for the ingested asset paths or test identifiers.
[EXTERNAL_DOWNLOADS]: The skill references the odiff-bin project on GitHub for image comparison during testing.
Evidence: Mentions https://github.com/nicolo-ribaudo/odiff-bin as the source for server-side image comparison binaries (SKILL.md).
[COMMAND_EXECUTION]: Provides example setup for Vite and Next.js plugins, as well as a standalone Node.js server using @editframe/dev-server.
Evidence: Includes TypeScript/JavaScript code for initializing dev servers that manage local file access and network requests (SKILL.md).
[CREDENTIALS_UNSAFE]: The skill references an API token but correctly guides the user to manage it via environment variables rather than hardcoding.
Evidence: Recommends using process.env.EF_TOKEN to initialize the API client (SKILL.md).

dev-server