audience-builder

Skill: Audience Builder

Two modes, one skill. The context determines which to run.

ICP mode (service business): Who will pay you? What do they care about? What language do they use? Persona mode (product builder): Who will use this? What job are they hiring it for? What does their day look like?

STEP 0 — DETECT MODE

READ data/2-Domaines/business-profile.json → revenue_model (service / product / both)

Context signal	Mode
"ideal client", "who should I target", "sales ICP"	ICP
"persona for my product", "user persona", "/solo:build" context	Persona
Both present or unclear	Ask: "Are you defining who will pay you (ICP for services) or who will use a product (persona)?"

---

ICP MODE

Step 1 — Data Classification

Data available	ICP type	Action
10+ past clients	Evidence-based	Build from facts
3–9 clients	Hypothesis ICP	Flag all assumptions
0–2 clients	Aspirational ICP	All assumptions — warn
B2B context	Add buying committee	5-role map required

Step 2 — Discovery Questions (don't generate without these)

Core:

• "Describe your best client in one sentence — not a job title, what makes them ideal?"
• "What problem did they have when they found you?"
• "Why did they really hire you? (not the stated reason — the underlying fear or desire)"
• "What exact words did they use to describe their problem?"

For B2B only:

• "Who else was involved in the decision?"
• "How long from first contact to signature?"
• "Who could have vetoed the purchase?"

Step 3 — B2B Buying Committee (B2B only)

Role	Objective	Fear	Influence
Economic Buyer	ROI, strategic fit	Wasted budget	Final sign-off
Technical Buyer	Integration, risk	Vendor lock-in	Gates shortlist
User Buyer	Ease of use	Change overload	Can kill adoption
Champion	Career advancement	Sponsors a failure	Drives case
Gatekeeper	Process, compliance	Liability	Can delay

Primary content target: usually Champion or User Buyer.

Step 4 — Output

Save to: data/2-Domaines/icp.json

Key fields: demographics, professional_profile, psychographics, pain_points (with real_why at level 4-5), language_patterns (verbatim from research), buying_triggers, b2b_buying_committee (if applicable).

Flag every field filled with assumption: [ASSUMPTION — validate with X clients]

---

PERSONA MODE

Step 1 — Research Approach

Mine before inventing:

• Reddit: r/[niche] + "frustrated" / "wish there was" / "why doesn't X"
• Product Hunt comments on competing products
• App store 1-star reviews of alternatives
• LinkedIn posts by target job title

Output verbatim quotes — never paraphrase audience language.

Step 2 — Persona Structure

# Persona: [Archetype Name]

## Snapshot
- Who: [Role, context, 1-sentence description]
- Quote: "[Something they'd actually say — from research if possible]"

## Jobs to Be Done
Primary JTBD: "When [situation], I want to [motivation], so I can [outcome]"
Secondary JTBDs: [list]

## Pain Points
- [Pain 1] — severity: High/Med/Low
- Source: [verbatim from research or "inferred"]

## Behavioral Signals
- [Workaround they currently use]
- [Tool they've tried and abandoned]
- [Community they're active in]

## Trigger Moments
[The specific life/work situation that makes them search for a solution — this is when to reach them]

## What "Done" Looks Like
[Their definition of success — specific and behavioral, not aspirational]

Save to: data/1-Projets/[product]/persona-[name].md

Step 3 — Persona Confidence

Interviews done	Confidence	Impact
0	Very Low — hypothetical	Flag every assumption
1–3	Low — early signal	Use cautiously
5–7	Medium — usable	Proceed with caveats
8+	High	Proceed with confidence

Always state: "This persona has [confidence level] confidence based on [N] user interviews."

---

Integration Points

• Replaces: icp-creator, proto-persona (both deprecated)
• Feeds: draft-outreach, proposal-generator, content-calendar, diagnostic-builder, discovery-synthesizer-agent
• Saves to: data/2-Domaines/icp.json (ICP mode) or data/1-Projets/[product]/persona.md (persona mode)