Skip to content

Ticker due-diligence agent

Goal: before your assistant says anything confident about a single asset, have it cross three independent views — the tape, the editorial corpus, and any breaking events — and flag divergences.

Tools: get_ticker_brief, get_ticker_research, get_ticker_sentiment, get_ticker_critical_events.

The pattern

"Do a full Coin Daily due-diligence pass on SOL-USD: the current brief, how it got here, what the corpus is saying, and any flagged events. Call out any disagreement between them."

A capable agent will:

  1. get_ticker_brief(SOL-USD) — current direction, conviction, key levels, developments.
  2. get_ticker_research(SOL-USD) — the 2h→168h arc: was today's read stable, or a fresh shift?
  3. get_ticker_sentiment(SOL-USD) — corpus attention + sentiment trajectory.
  4. get_ticker_critical_events(SOL-USD) — exploits, depegs, halts, regulatory actions.

Then it reconciles them.

The divergences worth catching

  • Clean tape, ugly news. Brief reads neutral/long_bias but critical_events flags an exploit — act on the event, not the tape.
  • Quiet price, rising attention. Flat brief but sentiment velocity spiking — something is building before it shows in the read.
  • Snapshot vs arc. A neutral, conviction 0.55 brief means one thing if it's been steady, another if conviction just fell from 0.7 — the research arc disambiguates.

Why it's better than unaided

The model can't see the publication's tape signal, its corpus, or its event flags. Crossing all four turns "here's a take on SOL" into "here's a take, and here's the one thing that contradicts it" — which is the part that protects the user.