The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/auth.mjs uses child_process.execFile to launch the system's default web browser (via open, start, or xdg-open) for the purpose of completing the OAuth 2.0 authorization flow.
[EXTERNAL_DOWNLOADS]: The script scripts/fetch.mjs performs HTTPS GET requests to official X API domains (api.twitter.com and api.x.com) to retrieve bookmark metadata and download media assets (images and videos).
[DATA_EXFILTRATION]: No unauthorized data exfiltration was detected. Sensitive credentials, including the Client Secret and generated OAuth tokens, are stored only on the local machine in .env and .tokens.json files with restrictive permissions (chmod 600).
[PROMPT_INJECTION]: The skill ingests untrusted data from the X API (tweet text) and writes it into local Markdown and text files. While these files are not directly fed back into the agent by the skill itself, they lack boundary delimiters, creating a potential surface for indirect prompt injection if a user manually provides the output files to an LLM in a subsequent session.

x-bookmarks-fetcher