The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the user to install two third-party packages from the NPM registry.
Evidence: npm install -g socialclaw@0.1.12 and openclaw plugins install npm:@xquik/tweetclaw@1.6.31.
The socialclaw package is the primary tool for the skill's functionality, while @xquik/tweetclaw is an optional plugin for retrieving X (Twitter) evidence packets. Both packages originate from non-trusted external sources.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from social media providers and local files which could contain malicious instructions.
Ingestion points: The skill reads social media account lists (socialclaw accounts list), post delivery status and analytics (socialclaw status), and campaign research input from the TweetClaw plugin (SKILL.md).
Boundary markers: No explicit boundary markers or instructions to ignore embedded commands are present in the processing logic described.
Capability inventory: The skill possesses the ability to perform network operations (publishing posts to 13 platforms), write to the local file system (schedule.json), and execute shell commands via the socialclaw and tweetclaw CLIs.
Sanitization: The instructions do not mention sanitization, validation, or escaping of the data retrieved from external social media signals before it is used to influence future scheduling or publishing actions.
[COMMAND_EXECUTION]: The skill relies heavily on the execution of CLI tools (socialclaw, openclaw) with various arguments, including file paths and JSON configurations, providing a wide command execution surface.
Evidence: Multiple command patterns such as socialclaw apply -f schedule.json, socialclaw assets upload --file ./image.png, and openclaw plugins install.

social-publisher