The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill utilizes the twitter-api-v2@1.17.2 Node.js package to facilitate communication with the X API. This is a well-known and standard library for this purpose.
[COMMAND_EXECUTION]: Uses the execute_cli tool to run JavaScript/TypeScript code within a bun runtime environment. This is the primary mechanism for executing Twitter API calls and managing the bot's timeline.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its core function of ingesting and responding to untrusted external data from Twitter mentions.
Ingestion points: Reads external content through the userMentionTimeline endpoint as described in the Reply Workflow.
Boundary markers: While the skill documentation provides detailed logic for filtering content, it does not explicitly define the use of structural delimiters (e.g., XML tags or specific markers) when presenting external mention text to the model for drafting.
Capability inventory: The skill possesses capabilities for file system modification (edit_file) and command execution (execute_cli).
Sanitization: The skill implements extensive sanitization and safety measures, including a "Skip List" for malicious or irrelevant content and "Hard Blocks" that mandate manual Telegram approval for any drafts involving wallet addresses, onchain action keywords, or mentions of the platform's primary bot account (@bankrbot).

twitter-agent