Positioning messaging

hs-positioning-messaging

Built on April Dunford's Obviously Awesome method. Output discipline from MKT1/Emily Kramer: deliver the answer, not the workings. Self-improving memory via Pawel Huryn's compounding loop.

---

Onboarding

Run this block once at the start of every new session. No output until complete. One question at a time. This is the Settings screen — not discovery.

Step 1 — Check brain:

Does /foundation/brain.md exist?
  YES → Load silently. Note ICP, alternatives, energy state. Proceed.
  NO  → Surface once: "No brain found. You can still run this skill, but output
         will be less precise. Run hs-product-marketing-context first for
         sharper results. Continuing."

Step 2 — Select mode: Ask: "What are you trying to accomplish today?"

Present options (let user pick one — no typing required):

• [BUILD] — New positioning from scratch
• [AUDIT] — Review and score existing messaging
• [FLETCH] — Internal positioning deck (6 slides)
• [SALES-ENABLEMENT] — Persona cards + competitive playbook
• [HOMEPAGE] — Production-ready hero, pillars, and CTAs

Step 3 — Confirm persona focus: Ask: "Which persona are we positioning for today?"

• If brain has ICP: surface it — "Based on your brain: [ICP]. Is this correct?"
• If no brain: ask for title, company type, and triggering situation

Step 4 — Confirm alternatives: Ask: "What are buyers comparing you against?"

• Minimum 3 required (direct, status quo, do-nothing)
• If brain has alternatives: surface and confirm

Step 5 — Surface any attached images: Parse all attached images before reading user text. Extract claims, flags, and implied category. Report as: "Your current materials say: [findings]"

After Steps 1–5: state the session setup back to the user in one paragraph. Await confirmation. Then proceed to Phase 1.

---

Trigger

• When: Positioning, messaging, value prop, sales copy, competitive messaging, homepage copy, messaging audit, or any request involving how a product should be described to buyers
• Not for: Product specification → use hs-product-requirement-doc. Buyer persona building → use hs-buyer-personas. Alternatives mapping in isolation → use hs-alternatives-map first.
• Example prompts:
  • "Build positioning for [product]"
  • "Our messaging sounds like everyone else"
  • "Help me position against [competitor]"
  • "Write our homepage headline and pillars"
  • "Audit this messaging for coherence"
  • "We can't explain what we do to sales"
  • "Is this positioning working?"

---

Inputs

• Args: Product/market context, target persona (required), alternatives (3+ minimum, must include status quo + do-nothing), customer evidence or win/loss signals
• Defaults: If no persona named, surface from brain or ask. If fewer than 3 alternatives, block and ask before proceeding.
• Context keys:
  • /foundation/brain.md — preferred. Sections 1 (product), 2 (ICP), 3 (alternatives), 5 (voice) if available
  • knowledge/rules.md — load at session start; apply confirmed patterns
  • knowledge/hypotheses.md — load at Phase 5; check for active tests
  • references/positioning-worksheet.md — load at Phase 3
  • references/example-messaging-doc.md — load at Phase 6 as quality bar
  • references/validation-playbook.md — load at Phase 6 or on request

---

Pre-flight

• Load /foundation/brain.md if it exists. Extract ICP, positioning hypothesis, competitive alternatives silently.
• If brain exists but Section 3 (Alternatives) is empty or marked 🔴 Placeholder: block and surface: "⚠️ Your alternatives map is incomplete. Run hs-alternatives-map first, then return here. I need named alternatives before positioning."
• If fewer than 3 alternatives provided by user and not found in brain: block intake and ask for status quo + do-nothing before proceeding. Hard stop.
• If no brain found: surface once, non-blocking. Continue.

---

Steps

All modes run all 6 phases unless noted. Shortcuts produce weak output.

Phase 1: Discovery

Run as a conversation. One question at a time. Never generate on the first message.

Hard floor — all four required before Phase 2:

• Product: what it does today (not roadmap)
• Buyer: one primary (title, company type, triggering situation)
• Alternatives: 3+ including status quo and do-nothing
• Outcome: single most important buyer result in one sentence

Ideal additions: T1 customer quotes · current messaging/screenshots · sales objections · win reasons · pricing model · company stage

Evidence tier on input:

• Verbatim quotes, win/loss transcripts, NPS verbatims → T1 (anchor a pillar)
• Usage data, churn rates, win rates, A/B results → T2 (support a pillar claim)
• Product descriptions, PMM assertions, founder hypotheses → T3 (must carry [T3 — NEEDS VALIDATION]; cannot headline a pillar)

Exit: Summarise all four hard-floor items back to user. Await confirmation.

---

Phase 2A: Market Energy Check

Identify primary segment energy state before mapping competitors.

State	Buyer situation	Messaging job
M1 — Potential	No solution; desire blocked by barriers	Remove barriers; unlock demand
M2 — Kinetic	Doing it badly (spreadsheets, manual)	Redirect effort to the better path
M3 — Captured	Already using a solution of this type	Win share on one specific dimension

Confirm with user. Primary positioning serves one state only. Multiple states require separate tracks — never mix in primary messaging.

---

Phase 2B: Competitive Mapping

Map all five alternative types honestly:

1. Direct competitors (same category)
2. Adjacent competitors (overlapping use case)
3. Status quo (spreadsheets, legacy tools, manual process)
4. Build in-house
5. Do nothing

For each: what it does well / what it fails / what the buyer must believe to choose it.

Find the gap — intersection of: buyer cares deeply + product does well today + no alternative addresses it.

Exit — cannot proceed without completing this sentence:

"For buyers who need [outcome], alternatives fail because [gap], and we address this with [specific capability today]."

If you cannot complete it: the product has a positioning problem messaging cannot fix. Stop and surface this to the user.

---

Phase 2C: Options-First

Generate 3–4 distinct positioning bets before committing to one direction. Each bet: the claim · target segment + energy state · primary risk · what it forces you to stop saying.

Present to user. Await selection. Do not proceed to Phase 3 without a chosen direction.

---

Phase 3: Positioning Development

Load references/positioning-worksheet.md. Follow it exactly.

Dunford sequence:

1. Competitive alternatives (from Phase 2B)
2. Unique attributes — specific only ("3-click onboarding" not "better UX")
3. Value translation — "So what?" × 3 levels; Level 3 belongs in messaging
4. Who cares most — narrow to trigger-event segment, not demographic
5. Market category — frame that makes strengths obvious, not odd
6. Positioning statement × 3 versions:

   "For [segment] who [situation], [product] is the [category] that [differentiator], unlike [alternative] which [limitation]."

Pick sharpest version. Exit: Statement cannot be said honestly by any named competitor. If it can, rewrite.

---

Phase 4: Messaging Hierarchy

Build top-down. Layer 3 claims must trace to Layer 1. Cannot trace → remove.

Layer 1 — Strategic Narrative (150–250 words) Old world → Shift → New world → Your role. Product appears after the problem is fully established.

Layer 2 — Core Positioning Tagline 5–8 words · One-liner 15–25 · Elevator pitch 50–75 · Boilerplate 75–100

Layer 3 — Key Message Pillars (exactly 3–4, MECE) Each: headline claim + 2–3 sentence explanation + 2–3 proof points

Layer 4 — Persona Cards (1–3 max) Each: pain in their language · promise tied to their metric · 3 proof points · one objection pre-handled verbatim · stage-appropriate CTA

Exit: All layers complete. Persona CTAs match actual sales motion.

---

Phase 5: Differentiation Stress-Test

Load knowledge/rules.md and knowledge/hypotheses.md. Apply confirmed rules first, then check hypotheses for active tests.

4-question test per differentiator. YES or NO only:

1. TRUE? — Provable with evidence a skeptical buyer accepts
2. RELEVANT? — Buyer cares enough to switch or pay more
3. UNIQUE? — 2+ competitors cannot honestly claim the same
4. SUSTAINABLE? — Still true in 12–18 months

Scoring (no exceptions):

• 4/4 → Layer 3 pillar headline
• 3/4 → supporting proof point only
• ≤ 2/4 → removed entirely

---

Phase 6: Output Packaging

Load references/example-messaging-doc.md. That is the quality bar.

Self-Verification Gate — binary pass/fail. Re-enter named phase on any failure. Do not deliver with a caveat instead of a fix.

Check	Pass condition	Fail →
Proof coverage	Every claim has T1/T2 or carries [T3] flag	Phase 4
Narrative coherence	Layers 1→4 read as one story	Phase 4
Competitor specificity	Statement cannot be said by any named competitor	Phase 3
Stress-test integrity	All pillar headlines 4/4	Phase 5
Persona count	≤ 3	Phase 4
Jargon	Zero: leverage, seamless, best-in-class, robust, turnkey, holistic, enterprise-grade, disruptive, synergy, game-changing, "powered by AI" as standalone claim	Fix inline
Vision flags	All [VISION FLAG] items addressed or documented as rejected	Positioning Traps section

Before delivering, state: "Self-verification passed. [N]/7 checks clear."

Mode outputs:

BUILD — All 4 layers + stress-test table + approved/forbidden language + channel usage guide + positioning traps rejected + validation plan

AUDIT — 5-dimension score (specificity / differentiation / coherence / proof / buyer language) + P1/P2/P3 rewrite queue + before/after for P1 items

FLETCH — 6 slides: segment + trigger · alternatives map · gap · positioning statement · pillars + proof · homepage wireframe with copy

SALES-ENABLEMENT — Persona cards (pain / promise / proof / objection / CTA) + competitive playbook with "never say" column

HOMEPAGE — Hero headline <8 words · subhead <20 words · 3 pillar headlines

• 1-sentence descriptions · primary + secondary CTA · 2 proof strip options. No placeholders.

All modes include: Approved language · Forbidden language + reason · Version + review date · Next steps from references/validation-playbook.md

---

Outputs

• Files written: n.v.t. — outputs delivered in chat only
• Chat output format: Mode-specific (BUILD / AUDIT / FLETCH / SALES-ENABLEMENT / HOMEPAGE). All modes include approved/forbidden language list, version, review date, and next steps
• External side effects: n.v.t.

---

Verification

• Positioning statement cannot be said by any named competitor
• All pillar headlines scored 4/4 on stress-test (true, relevant, unique, sustainable)
• Self-verification gate passed all 7 checks before delivery
• Mode-specific output matches expected format
• Evidence tier respected: T1/T2 claims anchor content; T3 claims flagged
• Onboarding confirmed mode, persona, and alternatives before Phase 1 started

---

Do Not Use For

• hs-product-marketing-context — when the task is buyer profiling or market definition → use that skill first
• hs-value-prop-statements — when you only need segment-specific value props without a full messaging hierarchy
• hs-competitive-battlecard — when building sales competitive response cards; use that skill after positioning is locked
• hs-icp — when the primary need is ICP definition, not positioning output
• hs-alternatives-map — when the primary need is mapping competitive alternatives in depth; use that skill first, then return here
• experiment-doc-builder — when testing messaging assumptions; this skill builds messaging, not validates it experimentally

---

Commands

/build [product name or context]

Run full BUILD mode: Onboarding → all 6 phases → 4-layer messaging document.

/audit [existing copy or landing page]

Run AUDIT mode: score existing messaging + rewrite queue.

/fletch [product name]

Run FLETCH mode: 6-slide internal deck + homepage wireframe.

/sales [product name]

Run SALES-ENABLEMENT mode: persona cards + competitive playbook.

/homepage [product name]

Run HOMEPAGE mode: production-ready hero + subhead + pillars + CTA.

/settings

Re-run the Onboarding block. Use to change mode, persona, or alternatives mid-session without losing prior work.

---

Learning Mode Close

At session end:

"Run Learning Mode close? I'll log what worked and update the skill's memory."

If yes, in order:

1. Check knowledge/rules.md — demote any rule contradicted this session
2. Check knowledge/hypotheses.md — promote any confirmed 3+ times; log new observations
3. Check decisions/INDEX.md — log any structural decision made this session
4. Append to sessions/log.md using the session format
5. Update knowledge/INDEX.md status counts if rules/hypotheses changed

No file is written without explicit user approval. No updates mid-session.

---

Reference Files

File	Load when
knowledge/INDEX.md	Session start. Always first.
knowledge/rules.md	Session start + Phase 5
knowledge/hypotheses.md	Phase 5 + Learning Mode close
knowledge/false-beliefs/catalog.md	AUDIT mode + diagnosing weak positioning
decisions/INDEX.md	Before any structural decision (category, segment, evidence tier)
references/positioning-worksheet.md	Phase 3
references/example-messaging-doc.md	Phase 6
references/validation-playbook.md	Phase 6 + on request
sessions/log.md	Learning Mode close only

---

Operating Rules

1. Run Onboarding before everything. Mode, persona, and alternatives must be confirmed before Phase 1 starts. No exceptions.
2. Load brain first. Context changes which energy state (M1/M2/M3) applies.
3. Parse attached images before asking discovery questions. Every explicit claim and visual signal tells you something.
4. No positioning without 3+ alternatives — status quo mandatory. If you can't name it, you don't understand the choice.
5. Never skip phases. All 6 phases run for every mode. Shortcuts produce weak output.
6. Primary persona must be ranked. Ask: who has this problem worst, right now?
7. Stress-test all pillars before delivery. 4/4 or remove. No exceptions.
8. Layer 3 claims must trace to Layer 1. Cannot trace = doesn't belong.
9. Evidence tier is binding. T3 claims must be flagged. Don't hide assumptions.
10. /settings resets Onboarding only. Prior session work is preserved.

---

Quality Gate

Runs after output generation, before delivery. Surface failures — do not deliver incomplete output.

Check	Standard	Pass =
Onboarding complete	Mode, persona, alternatives confirmed	Yes
Proof coverage	Every claim has T1/T2 or carries [T3] flag	Yes
Narrative coherence	Layers 1→4 read as one story	Yes
Competitor specificity	Statement cannot be said by any named competitor	Yes
Stress-test integrity	All pillar headlines 4/4	Yes
Persona count	≤ 3	Yes
Jargon free	Zero forbidden terms	Yes
Vision flags resolved	All [VISION FLAG] items addressed or rejected	Yes
Mode format correct	Output matches expected format for selected mode	Yes

---

Self-Improvement Loop

Before every session:

1. Load knowledge/INDEX.md — load knowledge/rules.md and knowledge/hypotheses.md
2. Check decisions/INDEX.md for any prior structural decisions relevant to this product/domain
3. Parse all attached images before reading the user's text
4. Run Onboarding block

After every session where output was produced:

1. Extract any positioning pattern → knowledge/positioning/patterns.md
2. If the same stress-test failure appears 3+ times → propose rule addition to knowledge/rules.md
3. If a competitor alternative was underestimated 3+ times → propose update to knowledge/competitive-blind-spots.md
4. Append session summary to sessions/log.md

Self-Improvement Trigger format — surface before encoding, never silently:

🔁 SELF-IMPROVEMENT TRIGGER
Pattern: [what was observed this session]
Proposed update: [exact wording of what would be added or changed]
Location: [file path]
Awaiting approval before encoding.

---

Changelog

v2.1.0 — 2026-06-15

Added Onboarding block (Ramp-style setup wizard) as a standalone section before Trigger. Added /settings command to re-run Onboarding mid-session. Updated Operating Rules to 10 rules (was 8) — added onboarding enforcement and settings reset rule. Added Quality Gate row for Onboarding completion check. Fixed Self-Improvement Trigger format to match SKILL-SPEC v2.0.0 surface-before-encoding requirement. Description expanded with additional trigger phrases for passive discovery cases. Evals file added at evals/evals.json.

Architecture decisions:

• Onboarding runs before Phase 1, not inside it — separates configuration from discovery
• /settings command preserves session work while resetting configuration — parity with how Ramp handles settings without losing context
• Trigger format block is now a standalone declared section at end of Self-Improvement Loop per spec requirement
• Onboarding block checks brain, sets mode, confirms persona, confirms alternatives, and parses images — 5 explicit steps that previously were implicit in Pre-flight and Phase 1

v2.0.0 — 2026-06-11

Full rebuild to SKILL-SPEC v2.0.0. Added all 7 required sections. 6-phase discovery flow formalized. Stress-test 4-question rubric. 5 output modes. Evidence tier classification. Operating rules + Quality Gate + Self-Improvement Loop.

v1.0.0 — [date]

Initial build. Phase-based positioning framework.