matteotitta Research source (Exa)
Default: Exa, per .claude/rules/exa-protocol.md (auto-loaded for research, audit, competitor, ICP, AEO, content sourcing, sales prospecting work).
Primary Exa tools for this skill: company_research_exa, web_search_exa.
Use case: pre-call account intel for prospect-prep.
Tool surface during the migration window:
- New plugin (preferred):
mcp__plugin_exa_exa__company_research_exa(afterclaude plugin i exa@claude-plugins-official). - Legacy MCP (still mounted):
mcp__exa__company_research_exa. - Both backends route to the same Exa API — they don't double-bill.
Citation: every Exa-derived claim uses [VERIFIED: exa_search, {url}, accessed {YYYY-MM-DD}] per .claude/rules/ontology.md.
Quality gate (research outputs): ≥3 sources per major claim, ≥50% [VERIFIED] confidence, date filter for any "recent / latest" claim, no fallback to WebSearch without flagging the data gap.
Worked examples + tool catalog: .claude/skills/meta-skills/exa/.
Client discovery
Generate tailored discovery call scripts based on prospect context. Maps research to targeted questions and listening cues for qualification.
Doctrine inherited (Step 7 — 0626 rollout, locked 2026-06-04)
Output complies with output-tenets.md, output-simplicity.md, outbound-research-hygiene.md. Step 6 calibration: see [[feedback_execution_doctrine_refinements_step6]].
Refinements applied: R1 (discovery script is internal rep-facing — inline cites stay for QA), R3 (question framing operator-direct), R9 (verb-led script section names).
Process at a glance
INPUT VALIDATION → RESEARCH PROSPECT → GENERATE QUESTIONS → ADD CUES → REVIEW & CHAIN
Three steps:
- Research prospect website + LinkedIn + sent materials, build opening playbook
- Generate tailored questions across 8 categories
- Add listening cues + qualification criteria
Full flowchart, step-by-step runbook, and self-evaluation gate live in the premium reference.
Claude Code triggers
Invoke this skill when user says:
- "Discovery call for [company]"
- "Discovery questions for [prospect]"
- "Prepare for discovery with [company]"
- "Qualify this prospect"
- "Discovery prep"
Do NOT invoke when:
- User wants company research only → Use
company-contextskill - User wants to create a proposal → Use
client-proposalsskill - User wants ICP research for a client → Use
icp-behaviouralskill
Inputs
Required
| Input | Description | Source |
|---|---|---|
| Company identifier | Website URL, name, or any context | User provides |
Optional (improve quality)
| Input | How it helps |
|---|---|
| company-context output | Pre-researched traction and qualification |
| Funding stage | Sets expectations for budget and scope |
| Known pain points | Focuses discovery questions |
| Meeting context | Any prior conversations or emails |
If inputs are missing: Ask for company URL or name.
Discovery question categories (8 total)
| Category | Purpose | Anchor question | Time |
|---|---|---|---|
| Company/product | Stage, GTM motion, marketing function | "How is your marketing function structured today?" | 5-7 min |
| Goals and priorities | Triggers, success metrics, past failures | "What triggered this conversation now?" | 5-7 min |
| ICP and competitive | Buyer personas, competitors | "Who are your top 3 competitors?" | 5-7 min |
| Messaging and positioning | Current frameworks, customer voice | "How would your best customers describe you?" | 5-7 min |
| Website and conversion | Satisfaction, scope, resources | "What's your conversion rate today?" | 3-5 min |
| Content and distribution | Founder time, LinkedIn priority | "How much time can founders dedicate to content?" | 3-5 min |
| Resources and constraints | Budget, team, decision-makers | "What's the budget range you're working with?" | 3-5 min |
| Timeline and next steps | Urgency, process | "When do you need to see results?" | 3-5 min |
Full question bank by category in the premium reference.
Anti-hallucination guardrails
- Base context on research. Use actual website content, not assumptions.
- Tailor questions to findings. Reference specific things you observed.
- Mark assumptions. If inferring from limited data, note explicitly.
- No invented details. If you can't find funding info, don't make it up.
Integration with other skills
| Skill | Relationship |
|---|---|
| company-context | Pre-research prospect before discovery |
| sales-call-playbook | Generate during-call guide (run immediately after discovery) |
| client-proposals | Create proposal after discovery call |
MCP data integration
Level: 0 — Context (heavy pulls)
Inherits from: company-context (if already ran for this company — skip Exa/Firecrawl/Apollo)
Pulls fresh
| Source | What to pull | Tool | When |
|---|---|---|---|
| Exa | Prospect company profile | company_research_exa |
Always |
| Firecrawl | Prospect website scrape | firecrawl_scrape |
Always |
| Apollo | Org + key people enrichment | apollo_search_companies, apollo_search_people |
Always |
| YouTube | Prospect YouTube presence/content | get_transcript |
If YouTube channel exists |
| Slack | Prior conversations with prospect | slack_search_public |
Always |
| Granola | Previous meetings with prospect | search_meetings |
Always |
Fallback (no MCP)
- WebSearch + WebFetch for prospect research
- Manual LinkedIn research for contacts
- Manual conversation history review
Final ship gate
Run /premortem --output before ship. See /premortem skill for the 5 execution domains (will-it-resonate / will-it-convert / will-it-stay-on-brand / will-stakeholder-push-back / will-it-degrade-over-time) and output template.
Trivial-case escape: ## Premortem\nNo failure modes — trivial change satisfies the contract for genuinely trivial outputs.