The Agent Skills Directory

[PROMPT_INJECTION]: The skill exposes a surface for indirect prompt injection by processing live scene data.
Ingestion points: Data enters the agent context through methods such as inspectEntity, inspectText, and inspectMarkdownStream (found in SKILL.md) which read application state and textual content.
Boundary markers: The documentation does not specify the use of delimiters or instructions for the agent to ignore potentially malicious commands embedded in the scene's text or metadata.
Capability inventory: The skill allows detailed state extraction, snapshotting, and event tracing, providing the agent with extensive access to application data.
Sanitization: While plugin isolation is mentioned, there is no evidence of content sanitization or filtering for natural language instructions within the processed scene elements.
[EXTERNAL_DOWNLOADS]: The skill references several external Node.js dependencies.
Referenced packages include @vectojs/devtools, @vectojs/core, and @vectojs/ui.
These are identified as vendor resources belonging to the skill author, vectojs, and represent standard development dependencies.
The library includes security features for its backend, such as origin enforcement for cross-document requests.

vectojs-devtools