The Agent Skills Directory

[DATA_EXFILTRATION]: The skill manages sensitive authentication cookies (X_COOKIES) required for account access. It includes explicit instructions and code-level warnings to prevent the accidental logging or echoing of these secrets. Network transmission is restricted to the target service's legitimate domains for functional purposes only.- [COMMAND_EXECUTION]: The skill utilizes a two-step confirmation process for all state-changing commands such as posting tweets, following users, or deleting content. The CLI script is designed to perform a dry-run by default, requiring a specific --confirm flag to be appended manually to execute the live action, which ensures full user oversight and prevents silent account modification.- [REMOTE_CODE_EXECUTION]: The skill includes a compatibility patch that fetches a public JavaScript file from the service provider's official domain (abs.twimg.com) to extract transaction indices. This process retrieves configuration data necessary for API interaction rather than downloading or executing arbitrary remote payloads.- [PROMPT_INJECTION]: Instructions within the skill reinforce safety protocols and user verification. Heuristic detections regarding concealment are determined to be false positives, as the skill's design specifically prioritizes transparency and manual verification for all user-impacting operations.

x