The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes shell commands including curl, jq, and date to perform API requests and process data. It specifically uses a ./secretcurl wrapper to securely pass the XAI_API_KEY to the X.AI API endpoint (api.x.ai) for gathering social media signals.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from search engines, social media posts, and RSS feeds, which creates an attack surface for indirect prompt injection.
Ingestion points: Data is ingested in Phase 1 — Gather through the WebSearch tool, the x_search tool (via the X.AI API), and by fetching XML content from URLs defined in memory/feeds.yml using curl.
Boundary markers: Absent. The instructions do not define specific delimiters or instructions to treat the fetched news content as data only, increasing the risk of the agent following instructions embedded in the articles or posts.
Capability inventory: The skill has access to network tools (curl, secretcurl, WebFetch), search capabilities (WebSearch), and a notification mechanism (./notify).
Sanitization: Absent. While the skill includes a filtering phase (Phase 2 — Filter) to ensure signal quality and relevance, it does not implement security-focused sanitization or escaping of the ingested text before it is processed by the agent.

digest