| name | twitter-browser-research |
| description | Use Twitter/X tooling to inspect search results, threads, profiles, and market commentary from analysts, traders, and company watchers. |
| metadata | {"marketbot":{"emoji":"🐦","triggers":["twitter","x search","tweet thread","fintwit","twitter sentiment"],"output":"twitter-browser-research-report","risk":"medium","freshness":"live","tools":["twitter_cli","browser_site"],"required_tools":["twitter_cli"],"alternative_required_tools":["browser_site"],"markets":["global","mixed"],"asset_classes":["equity","crypto","commodity","macro","etf"],"task_type":"browser-research","determinism":"tool-backed","priority":84}} |
Twitter Browser Research
Use this skill when the user needs X/Twitter-native market commentary, thread
search, or fast-moving social discussion around an asset, theme, or event.
Workflow
- Prefer
twitter_cli when available.
- search:
twitter_cli(operation="search", query="$NVDA guidance earnings revenue", search_type="Latest", max_count=12, exclude=["replies","retweets"], do_filter=true, min_likes=2)
- thread/detail:
twitter_cli(operation="tweet", target="<tweet-url-or-id>", max_count=30)
- user:
twitter_cli(operation="user", screen_name="@sama")
- user posts:
twitter_cli(operation="user_posts", screen_name="@sama", max_count=20)
- article:
twitter_cli(operation="article", target="<article-url-or-id>")
- If
twitter_cli is unavailable, use browser_site with Twitter/X adapters that exist in the runtime catalog. Prefer exact adapters such as:
twitter/search
twitter/thread
twitter/user
- Read references/adapter-examples.md when you need concrete adapter call patterns or fallback behavior.
- Focus on:
- recurring narratives
- analyst or trader commentary
- whether sentiment is accelerating or reversing
- Pair with
sentiment-analysis when a weighted conclusion is needed.
Rules
- Prefer
twitter_cli over browser_site for Twitter/X search, tweet detail, profile lookup, and user-post analysis because it exposes richer structured output.
- For ordinary narrative or sentiment requests, force
search_type="Latest", prefer cashtags for obvious equity tickers, add exclude=["replies","retweets"], and keep the search window small enough to reduce spam and engagement bait.
- If the query contains broad finance words like
guidance, outlook, or forecast, tighten it with earnings context such as the ticker cashtag, earnings, or revenue instead of searching the bare word alone.
- For ordinary analysis requests, do not call
status first. Only call status after a prior twitter_cli call reports an authentication or access failure.
- Start with one focused search, not a broad crawl. Only run a second search when the first query is too noisy or misses the core catalyst wording.
- If
twitter_cli returns structured ok: true data, stop there and synthesize the answer from that data. Do not switch to exec, browser_site, or local cache inspection for the same request.
- Use
browser_site only when twitter_cli is unavailable or returns an actual access failure for the requested read path.
- Do not guess undocumented
twitter/* adapters. If the catalog does not expose the adapter you want, say so and use the closest listed adapter instead.
- Treat Twitter/X as fast signal and distribution context, not verified fact by itself.
- Separate original reporting from repeated hot takes or engagement bait.