---
name: gather-info
description: Gather structured intelligence on a topic — thesis, debates, evidence, counterfactual, unknowns. The CIO's first-pass research verb.
visibility: public
type: draft
director: CIO
base: draft
url: navigateengine.com/skill/gather-info
---
# Gather Information
**The CIO's intelligence-gathering verb.** Take a question and return a structured, sourced briefing that names what is known, what is debated, and what is unknown. The goal is not the longest answer — it is the answer that lets the reader hold the full epistemic landscape in their head in one minute.
---
## input
A topic, a question, or a brief. Examples:
- `"How is the residential construction sector in NSW shifting in 2026?"`
- `"What are the live options for measuring AI-generated content in Google Analytics 4?"`
- `"Brief me on [competitor]'s product strategy over the last 18 months."`
If no topic is given, ask the user **one** sentence of clarification, then proceed.
---
## process — ASK → PREFLIGHT → DRAFT → AUDIT
1. **ASK** — restate the research question in one sentence. State scope boundaries (`in scope:` / `out of scope:`). One round only.
2. **PREFLIGHT** — emit the output skeleton (section names + content type per section) **before writing**.
3. **DRAFT** — fill the skeleton in one pass. Every numeric or named claim carries either a citation or the literal token `[UNVERIFIED]`.
4. **AUDIT** — run the contract below. If any item fails, revise. Show the audit table inline at the bottom of the deliverable.
### Steps inside DRAFT
1. **Restate** the research question (1 sentence) and **scope** (`in scope:` / `out of scope:`).
2. **Background** — what happened, when, why it matters. Each milestone gets a date.
3. **Current state** — what is true right now, in 3–5 numbered claims, each citation-bearing.
4. **Debates** — at least two opposing positions, each named with its strongest single argument and its strongest single counter.
5. **Counterfactual** — the single condition under which the current state would not hold.
6. **Sources** — 5–10 sources, each cited inline by `[n]` and listed in a Sources block at the bottom.
7. **Unknowns** — what you could not resolve, and what would resolve it.
---
## output
A single self-contained HTML file in NavigateEngine **B&W** styleguide (Geist Sans / JetBrains Mono optional, system stack acceptable; ink `#111`, paper `#fff`, border `#e5e5e5`).
### inline skeleton
```html
{{ title }}
{{ the condition under which the opposite recommendation is correct }}
Sources
{{ [n] author · publisher · date · URL }}
```
---
## contract — pre-commitment, not post-hoc
Read these before drafting. The output must satisfy every line; the bottom of the deliverable must show this table with PASS/FAIL + evidence.
- Research question restated as a single sentence.
- Scope boundaries stated (in / out).
- Background includes specific dates, not "recently" / "lately" / "now".
- ≥ 2 opposing positions, each with an argument AND a counter.
- Counterfactual is a single specific condition, not a vague caveat.
- 5–10 sources cited; every numeric or named claim carries `[n]` or `[UNVERIFIED]`.
- Inversion field is present and specific.
- Unknowns block is non-empty (if everything is known, the question was too narrow — say so).
---
## verify hook
If the briefing contains ≥ 5 numeric claims, invoke `/verify` against the draft before final delivery and append the verdict matrix to the deliverable. The CIO role module enforces this rule globally — this skill respects it.
---
## inversion (CIO-specific)
Confirming what the user already believes is failure. If your synthesis lines up cleanly with the user's framing, you have either missed the counter-position or the counter-position is genuinely weak. Say which.