> ## Documentation Index
> Fetch the complete documentation index at: https://docs.gdeltcloud.com/llms.txt
> Use this file to discover all available pages before exploring further.

# brief-output

> BLUF / inverted-pyramid analyst brief format with per-evidence-type citation templates and gdelt-event fenced blocks

## Overview

Use the `brief-output` skill when the user asks for a brief, an analytical write-up, a memo, a "what should I know about X" report, or a synthesis across multiple surfaces.

For lookups, schema help, extraction, comparison, or procedural tasks, do not force this template. Use the smallest clear shape (a sentence, a list, a table) that fits.

## Structure (BLUF / inverted pyramid)

```
🧭 Bottom Line
⚡ So What
👀 Watch Next
📊 Evidence and Signal
🔎 Details
```

The reader must be able to stop at any section break and still walk away with the analysis.

### 🧭 Bottom Line

2–4 bullets. Each states judgment, magnitude, duration, and confidence (qualitative — low / moderate / high). Lead with the most consequential claim.

### ⚡ So What

Practical implications. If the brief is for a known audience (founder, investor, policymaker), tailor the implication.

### 👀 Watch Next

Leading indicators or decision triggers. Specific, observable, and timed. "Watch the next CPI print on May 13" beats "watch inflation."

### 📊 Evidence and Signal

The returned data that supports the judgment. Lead with structured GDELT Cloud findings — a small table of the most-significant Stories or Events:

| Date | Story title (linked) | Significance | Magnitude | Why it matters |
| ---- | -------------------- | ------------ | --------- | -------------- |

Then layer in supporting surfaces: macro-finance with provider-resolved symbols and time-series moves, prediction-market probabilities with liquidity caveats and `public_url`, and web research only as enrichment.

### 🔎 Details

Caveats, sourcing limits, alternative interpretations, counter-signals.

## Citation templates by evidence type

| Surface            | Template                                                                    |
| ------------------ | --------------------------------------------------------------------------- |
| GDELT Cloud Story  | `[Story title](https://gdeltcloud.com/stories/<slug>)`                      |
| GDELT Cloud Entity | `[Entity name](https://gdeltcloud.com/entities/<slug>)`                     |
| Source article     | `[Domain — Title](article URL)` from `top_articles` or `get_story_articles` |
| Macro-finance      | `<symbol> <metric> <value>, <date> (Alpha Vantage)`                         |
| Prediction market  | `[Market title](public_url)`                                                |
| Web research       | `[Source — Title](URL)`                                                     |

**Quote the GDELT Cloud structured metric in the prose, not just the source.** "The Iran/TEE-01B story (significance 0.09, max\_linked\_event\_significance 0.40) is a low-volume cluster with a high-significance event — important but not yet trending" beats "this story is important." The metric *is* the analysis.

## Highlighting individual events

When 1–3 specific incidents anchor the analysis, render them as fenced `gdelt-event` blocks:

````
```gdelt-event
type: cameoplus
domain: TECHNOLOGY
event_type: SPACE OR SATELLITE EVENT
description: Iran allegedly used Chinese spy satellite to target US bases
country: —
date: 2026-04-15
actors: [Iran (IRGC), China (Earth Eye Co)]
magnitude: 2
goldstein: 0
market: 0.15
systemic: 0.55
fatalities: 0
url: https://gdeltcloud.com/stories/iran-allegedly-used-chinese-spy-satellite-to-target-us-bas-4917ab01
```
````

Pull fields from the returned Event card. If a field is missing, omit the line.

## Window disclosure

Always state the time window queried and the surfaces used:

> *Window queried: 2026-04-06 to 2026-05-06 (30 days). GDELT Cloud first; web research used for institutional context and corroboration.*

## Confidence calibration

| Level        | Criteria                                                                                  |
| ------------ | ----------------------------------------------------------------------------------------- |
| **High**     | Multiple independent sources, structured data corroborates narrative, no major gaps.      |
| **Moderate** | Single primary source with corroborating context, or structured data with narrative gaps. |
| **Low**      | Single source, contested claims, no structured data backing the claim.                    |

Don't hide behind "moderate" everywhere.

## Formatting rules

* Emojis only in section headers and very sparse labels — never one per bullet.
* Markdown headers short and consistent.
* Compact tables over long paragraphs.
* Source links as readable Markdown labels: `[Reuters](https://...)`, not bare URLs.
* Don't end the brief by asking if the user wants a follow-up. If you don't have enough to answer, say so up front and pose a single specific clarifying question.