whathappened logo

whathappened

Agent skill: adaptive X-only briefing of what happened + public opinion + debates

AI AgentsMIT 205 14Health 82
Grok Build onlyAgent Skills packageX @kunchenguiddiscord
Website GitHub

What is it?

What it is

An Agent Skill for Grok Build that aggregates public X (formerly Twitter) conversations into a concise, neutral briefing, including what happened, where the conversation is, opinion maps, live debates, and key posts.

Why it exists

To give users a single, up‑to‑date snapshot of a topic on X without relying on external web search, ensuring the briefing stays X‑centric, fresh, and opinion‑aware.

Who should use it

Grok Build users who want quick, neutral briefings on trending X topics, especially developers and agent builders needing real-time sentiment analysis.

Who should avoid it

Users without Grok Build access, those seeking multi-platform research, or anyone needing deep historical analysis beyond X.

How it works

A quick walkthrough in plain English

How whathappened works

Step 1 of 3

You interact with it

Open whathappened, send a request, or connect it to your stack.

Features

X-first public conversation aggregation
Adaptive window based on story velocity
Opinion map with camps and debates
Thread fetching for official and community posts
Entity resolution via X user search
Structured markdown briefing output
Integration with Grok Build native X tools
Optional web lookup for entity resolution
Ranking and clustering by engagement, freshness, and on-entity
Multiple modes: Breaking, Same-day, Story, Background

Advantages

  • Provides concise, neutral briefing from real X data
  • Prioritizes freshness with adaptive window
  • Includes rich context: opinion map, debates, receipts
  • No reliance on external web search for content
  • Leverages Grok Build’s native X stack for accurate data
  • Structured output suitable for agent consumption
  • Automatically resolves entity identity
  • Supports multi-thread analysis for comprehensive coverage

Disadvantages

  • Requires Grok Build native X tools; not usable on other agents
  • Skill refuses to run if X tools are missing; no fallback
  • Limited to X/Twitter content only
  • Dependent on X API availability and rate limits
  • Not a multi-harness research product; specific to Grok Build
  • Requires installation into Grok environment
  • Potential bias due to English/AI-Twitter skew
  • Day‑0 recency bias on announcements
  • No support for non‑X sources

Installation

FAQ

What is whathappened and what does it produce?

whathappened is a Grok Build Agent Skill that turns the X firehose around a topic into a short neutral briefing: what happened, where the conversation is, an opinion map, live debates, and receipts (real post links).

How do I install whathappened for use in Grok Build?

Install globally with `npx skills add kunchenguid/whathappened -g` (recommended) or locally with `npx skills add kunchenguid/whathappened`. You can also clone the repo and copy or symlink the skill folder into `~/.grok/skills/whathappened`.

Can I use whathappened outside of Grok Build?

No. The skill depends on Grok Build’s native X tools (`x_keyword_search`, `x_semantic_search`, `x_thread_fetch`, `x_user_search`). If those tools are missing, the skill should refuse rather than fake an X briefing from web search. Installing into other agents like Claude Code, Cursor, or Codex will not give a working run.

What kind of output does the skill generate?

The skill produces a markdown briefing with sections: Window, X sample, Entity resolve, What happened, Where the conversation is, Public opinion map, The live debates, Notable posts (receipts), and Gaps/caveats. It includes rough opinion‑camp shares, steelman vs critique points, and links to real X posts.

Are there any limitations or caveats I should be aware of?

Yes. The briefing can suffer from day‑0 recency bias, relies on an X/Twitter‑skewed sample, and may miss non‑English conversation. Weights or product details that are forthcoming (e.g., open weights promised for a future date) are treated as forward‑looking. The skill uses at most one web lookup solely to resolve the entity’s identity; all sentiment and discussion analysis stays X‑only.

How does the skill decide the time window for its analysis?

It uses an adaptive window: breaking stories are analyzed over minutes or hours, quieter topics widen only as needed to explain the story. The pipeline first measures velocity via a pulse (Latest) search, then commits to the shortest window that still captures the story—mode options are Breaking (hours), Same‑day, Story (days), or Background (longer only if needed).

Loading documentation…
View on GitHub

Featured in Videos

YouTube tutorials and walkthroughs for whathappened

Alternatives

Similar projects ranked by category, topics, and text overlap.

Compare
whathappened | MushyBook