The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/cookies.ts spawns a local browser process (Google Chrome, Microsoft Edge, or Chromium) using node:child_process.spawn. This is used to capture session cookies via the Chrome DevTools Protocol (CDP) when necessary credentials are not provided via environment variables.
[EXTERNAL_DOWNLOADS]: The skill fetches content from X's GraphQL API and downloads JavaScript bundles from abs.twimg.com to resolve API operation IDs. Additionally, scripts/media-localizer.ts downloads images and videos from pbs.twimg.com and video.twimg.com to the local filesystem for offline markdown reference.
[CREDENTIALS_UNSAFE]: A hardcoded Bearer token is present in scripts/constants.ts for guest access. The skill also manages sensitive session cookies (auth_token, ct0) which are persisted in a local cookies.json file in the application's data directory.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted text from external tweets and articles. This content is converted into Markdown and presented to the agent without advanced sanitization for embedded instructions.
Ingestion points: Content is retrieved from X via scripts/graphql.ts and scripts/thread.ts.
Boundary markers: The output is structured with YAML front matter and Markdown headers, but lacks explicit 'ignore instructions' delimiters for the ingested text.
Capability inventory: The skill has the ability to write files to the disk and spawn system processes (browsers).
Sanitization: Limited to basic Markdown escaping for media metadata.

baoyu-danger-x-to-markdown