The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to install an external plugin (@xquik/tweetclaw) from the NPM registry. This introduces third-party code into the execution environment that does not originate from the skill author or a predefined trusted organization.
[COMMAND_EXECUTION]: The instructions include shell commands for the user or agent to execute, specifically openclaw plugins install and openclaw plugins inspect. These commands facilitate the installation and runtime verification of external code.
[PROMPT_INJECTION]: The skill defines workflows for scraping tweets, replies, and mentions from X/Twitter. This data ingestion creates a surface for indirect prompt injection, as malicious text within scraped content could attempt to influence agent behavior. The skill includes a mitigation by requiring explicit human confirmation before performing state-changing actions like posting or exporting data.
[DATA_EXFILTRATION]: The skill is capable of exporting sensitive user data such as direct messages and follower lists to external domains (e.g., docs.xquik.com). While it mandates authorization and confirmation, the potential for data access and transfer to non-whitelisted domains is noted.

tweetclaw-twitter-automation