The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/cookies.ts launches a browser instance (Chrome, Edge, or Chromium) using child_process.spawn to allow the user to log in and capture session cookies. This process uses a dedicated, isolated profile directory within the skill's data folder to avoid accessing the user's primary browser data.\n- [DATA_EXFILTRATION]: Sensitive session cookies (auth_token, ct0, etc.) are captured from the browser and stored in a local cookies.json file as implemented in scripts/cookie-file.ts. These credentials are used to authenticate subsequent GraphQL requests to X.com.\n- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes untrusted content from X.com.\n
Ingestion points: External tweet and article content is ingested via the GraphQL API in scripts/graphql.ts.\n
Boundary markers: No delimiters or safety instructions are provided to the agent when it receives the formatted markdown output.\n
Capability inventory: The skill can execute commands (browser launch), write to the local filesystem (saving markdown and media), and perform network operations.\n
Sanitization: Content is formatted into markdown, but the body text is not sanitized to remove or neutralize potential instructions hidden in the tweet content.\n- [EXTERNAL_DOWNLOADS]: The scripts/media-localizer.ts module uses the fetch API to download images and videos from remote URLs discovered within the processed content and saves them to the local filesystem.

baoyu-danger-x-to-markdown