---
name: research-prospect
description: Dedicated prospect research protocol for hooks, context, and personalization.
visibility: internal
---

# Research Prospect (Dedicated)

Use this subskill for lead/prospect research before message drafting.
This protocol is standalone and does not depend on the generic `research` subskill.

## Objective

Produce high-signal prospect context for outreach:

- What they care about (posts/themes)
- Company context and momentum
- Personalization hooks and conversation starters

## Inputs

Provide as many as available. Use `"Unknown"` when missing.

- `name`
- `title`
- `companyName`
- `companyDomain`
- `companyLinkedinUrl` (optional)
- `linkedinUrl` (optional)
- `experienceSummary` (optional)

## Execution Backend Routing

Choose backend once:

1. If product-native MCP/tool parallel calls are available: batch independent
   retrieval units in one turn.
2. If the user explicitly asked for agent fan-out and `Task` is available: use
   Task subagents in parallel.
3. Else if `multi_tool_use.parallel` is available: batch independent tool calls via `multi_tool_use.parallel`.
4. Else: run units sequentially.

Never claim Task subagents were used when they were not.

## Evidence Retrieval Routing (Web)

Use host-native web tools for external context:

- Claude host: use `WebSearch` + `WebFetch`.
- Codex host: use `web.search_query` + `web.open`.

Hard rule:

- Do NOT use shell scraping for research extraction (`curl`, `_next` bundle grep, JS chunk scanning).
- If a page is JS-heavy and unreadable, pivot to alternate sources and explicitly mark what remained unavailable.

## Research Units (6)

Run these units with maximal safe parallelism for the selected backend.

1. Post Analyzer

- Use LinkedIn posts when `linkedinUrl` exists.
- Extract top themes, recent focus, and best hooks.

2. Homepage Snapshot

- Fetch company homepage when domain exists.
- Extract positioning/proof hints.

3. Company LinkedIn Snapshot

- Resolve company LinkedIn URL if needed, then call `fetch_company`.
- Extract company context and headcount hints.

4. Role Context (No Tools)

- Infer likely priorities, pains, and decision context from role/title.

5. Jobs + Funding Signals

- Search hiring/funding/press momentum.

6. Reviews + Traffic Signals

- Search review and traffic indicators.

## Output

Return exactly this structure:

```markdown
## {Name} - {Title} at {Company}

**Top of Mind:** [...]
**Company Context:** [...]
**Pain Points:** [...]
**Hooks to Use:** [...]
**Growth Signals:** [...]
```

If data is missing, use `None found` rather than guessing.