The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external social media posts and feeds them to an LLM to generate automated engagement, creating a surface for injection attacks.
Ingestion points: The agent ingests data by browsing home feeds, searching for niche keywords, and visiting influencer profiles as described in SKILL.md and references/browser-scripts.md.
Boundary markers: The provided documentation does not specify the use of boundary markers or specific instructions for the LLM to ignore embedded commands within the ingested tweets.
Capability inventory: The skill possesses significant automated capabilities including posting replies, creating original posts, following accounts, and liking content via Puppeteer (src/algorithmBuilder.js).
Sanitization: There is no mention of sanitizing or filtering external content before it is processed by the LLM (OpenRouter) for content generation.
[COMMAND_EXECUTION]: The skill relies on browser-based automation and headless Chrome control to perform its functions.
The skill documentation instructs users to execute JavaScript code directly in the browser's developer console (algorithmTrainer.js, thoughtLeaderCultivator.js), which requires access to the user's active authenticated session on x.com.
The automated engagement engine uses Puppeteer to simulate human behavior, including typing, scrolling, and navigating, which involves programmatically controlling browser actions based on external triggers.

algorithm-cultivation