The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the chrome-devtools MCP tool to perform browser-level actions such as new_page, click, and type_text. These operations are limited to the scope of interacting with x.com to compose and publish posts.
[DATA_EXFILTRATION]: The skill uses take_snapshot() to capture the browser's state for UI navigation and confirmation. While this involves reading session-specific data from the webpage, the instructions do not contain patterns for transmitting this data to unauthorized external endpoints.
[PROMPT_INJECTION]: As the skill adapts user-provided content into threads, it is technically susceptible to indirect prompt injection. However, the skill explicitly mandates that the agent must 'show the draft and ask for confirmation before clicking the publish button' and 'publish only when the user explicitly asks to post', which serves as a human-in-the-loop safety boundary.
[EXTERNAL_DOWNLOADS]: The skill references the chrome-devtools MCP tool as a dependency. No suspicious third-party package installations or remote script executions (e.g., curl | bash) are present.

x-thread-poster