The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill defines a workflow for processing untrusted external data (e.g., tweet text, trending topics) and interpolating it directly into Grok AI prompts.
Ingestion points: The skill ingests data from external tweets via the {tweet_text} placeholder and from Grok's own responses using the scrapeResponse() method in SKILL.md.
Boundary markers: The instructions do not include the use of delimiters (like triple backticks or XML tags) or specific instructions to the AI to ignore embedded commands within the ingested text.
Capability inventory: The described scripts (src/grokIntegration.js) automate browser interactions, including prompt submission and response scraping, which could be exploited if an attacker-controlled tweet contains malicious instructions that Grok follows.
Sanitization: The documentation provides no guidance on sanitizing or validating external content before it is interpolated into prompts or used in downstream content creation via threadComposer.js.

grok-ai