These prompts match Xcatcher's narrow contract: recent public posts from named handles. Each one tells the agent to preflight for free, obtain approval before spending, preserve source links and report incomplete coverage.
Replace the example handles with your own list. Xcatcher does not perform keyword-wide search.
Use Xcatcher to preflight @openai, @AnthropicAI and @GoogleDeepMind in normal mode. Show the modeled cost before any payment. After approved retrieval, list product announcements in time order, preserve tweet_link for every claim, and report any handle with no posts or an upstream failure.
Fetch a recent public-post snapshot for these company and founder handles: [HANDLES]. Group concrete launches, pricing changes, partnerships and hiring signals. Separate direct statements from your inference. Include coverage and source links.
Preflight these researcher handles: [HANDLES]. Start with normal mode. If the recent snapshot lacks enough context, explain why deep mode may help and show its live cost before requesting approval. Summarize papers, talks, datasets and technical debates with links.
Using my existing Xcatcher API key, collect these handles for today's digest: [HANDLES]. Use a stable idempotency key for this date and watchlist. Wait for the same task instead of recreating it. Return new themes, notable links, empty handles and per-handle failures.
The Skill supplies these decisions even when the user's prompt is shorter.
Call get_service_info, then preflight_crawl. Do not create a quote merely to calculate cost.
Choose normal for recurring snapshots. Use deep only when more per-handle context justifies higher cost and latency.
Show exact live terms before approval and keep wallet secrets, signatures, API keys and task tokens out of prompts.
Preserve tweet_link, distinguish source statements from inference and expose partial coverage.
No account or payment is required for either call.
Connect https://xcatcher.top/mcp/ and call preflight_crawl, or use POST /api/v1/preflight.
Call get_sample_result or open /api/v1/demo. Rows are synthetic and clearly labeled.