The Agent Skills Directory

[SAFE]: The skill provides legitimate browser automation functionality for the Jitter.video platform. All external domains (jitter.video, cloudfront.net, s3.amazonaws.com) are official components of the targeted service.
[INDIRECT_PROMPT_INJECTION]: The skill exposes an attack surface by allowing the agent to replace text and asset URLs within a web page based on user-provided data. While it lacks explicit sanitization, this is a standard capability for browser automation tools and is managed by the agent's internal safety protocols. 1. Ingestion points: replaceText and replaceAssetUrl functions in SKILL.md and utils/actions.ts. 2. Boundary markers: None explicitly defined. 3. Capability inventory: page.evaluate for browser interaction and fs.writeFileSync for saving rendered frames to a local temporary directory. 4. Sanitization: Not present in the utility scripts.
[DYNAMIC_EXECUTION]: The skill uses page.addInitScript to inject jitter-utils.js into the browser session and page.evaluate to execute commands. This is the intended design for providing programmatic control over the web application.

jitter