@ojesusmp/marketing-council 1.1.0 → 2.0.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (3) hide show
  1. package/README.md +61 -42
  2. package/SKILL.md +216 -571
  3. package/package.json +4 -4
package/README.md CHANGED
@@ -1,83 +1,102 @@
1
1
  # marketing-council
2
2
 
3
- > 12-seat marketing/sales/copy council with Alex Hormozi as 13th-seat coordinator and a Skeptic post-verdict challenger. Produces an integrated playbook for digital products, AI agents, web pages, courses, and indie services.
3
+ > Adversarial 3-round marketing review with discipline foreman. Pitcher argues the asset works. Skeptic argues it fails with a cheapest disconfirming test under 72 hours. Hormozi judges round 3 with SHIP / EDIT / KILL. Foreman applies four discipline checks and ships an amended edit list. Output target under 1,700 tokens.
4
4
 
5
- A Claude Code skill modeled on `council-of-12` and `forge-council`, specialized for selling. Twelve marketing disciplines analyze the product, Hormozi integrates via the value equation, and a Skeptic seat challenges the signed playbook to kill confirmation bias.
5
+ A Claude Code skill for asset-level critique: page, copy, offer, hook, ad, email. Replaces the prior 12-seat panel architecture with named adversaries to kill diffuse responsibility, filler output, fake-distinct headlines, and recommendations for assets that do not exist.
6
+
7
+ ## What it produces
8
+
9
+ A single markdown verdict with:
10
+
11
+ 1. **Substrate echo** — live? customers? proof assets? (3 questions, mandatory)
12
+ 2. **3 rounds of Pitcher vs Skeptic** — claim, mechanism, evidence, edit, counter, failure, test
13
+ 3. **Hormozi raw verdict** — SHIP / EDIT / KILL, max 5 numbered edits, restated Skeptic test, highest-leverage move
14
+ 4. **Foreman pass** — 4 discipline checks (surfaced assumptions / minimum edits / surgical traceability / verifiable test threshold)
15
+ 5. **Foreman-amended final edit list** — KEEP / NARROWED / SEQUENCED / DEFERRED labels per edit
16
+
17
+ Token budget: ~1,600 target, 1,700 hard cap.
6
18
 
7
19
  ## Install
8
20
 
9
- ### Option 1: git clone
21
+ ### Option 1: npm (recommended)
22
+
23
+ ```bash
24
+ npx @ojesusmp/marketing-council
25
+ ```
26
+
27
+ ### Option 2: git clone
10
28
 
11
29
  ```bash
12
30
  git clone https://github.com/ojesusmp/Marketing-Council ~/.claude/skills/marketing-council
13
31
  ```
14
32
 
15
- ### Option 2: download SKILL.md only
33
+ ### Option 3: download SKILL.md only
16
34
 
17
35
  ```bash
18
36
  mkdir -p ~/.claude/skills/marketing-council
19
37
  curl -fsSL https://raw.githubusercontent.com/ojesusmp/Marketing-Council/main/SKILL.md -o ~/.claude/skills/marketing-council/SKILL.md
20
38
  ```
21
39
 
22
- After install, restart Claude Code (or run `/skills reload` if your harness supports it). Invoke with `/marketing-council "<product to sell>"`.
23
-
24
- ## What it produces
25
-
26
- A single integrated playbook with:
27
-
28
- 1. **Positioning** — brand promise, enemy, leveraged Cialdini principles, multi-year story arc
29
- 2. **Hook & Headlines** — 3 distinct angles (pain / outcome / curiosity) + CTA + risk reversal
30
- 3. **Channel Mix** — priority order WITH stated reason
31
- 4. **SEO Foundation** — primary keyword + 3-5 long-tail + title tag + meta description
32
- 5. **Funnel + Email** — landing page structure + welcome sequence + launch sequence
33
- 6. **PR / Community** — 3-5 named outreach targets + founder voice + spectacle plan + crisis-response template
34
- 7. **Offer Twist** — Hormozi value-equation lens applied
35
- 8. **30-Day Launch Sequence** — week-by-week
36
- 9. **Tensions Log** — where seats disagreed and how Hormozi resolved
37
- 10. **Hormozi's Sign-Off** — offer in one sentence + 4x-value verdict + single highest-leverage move
38
- 11. **Skeptic's Counter-Argument** — load-bearing assumption + cheap disconfirming test + failure signal + clarification questions
40
+ After install, restart Claude Code. Invoke with `/marketing-council "<asset to review>"`.
39
41
 
40
42
  ## Invocation patterns
41
43
 
42
44
  ```
43
- /marketing-council "sell my AI agent to coders"
44
- /marketing-council "launch my SaaS landing page"
45
- /marketing-council "promote my new course"
46
- /marketing-council --web "TikTok hook for current trend"
45
+ /marketing-council "https://truepoint.example/agents"
46
+ /marketing-council "review my landing page"
47
+ /marketing-council "audit this offer: <pasted terms>"
48
+ /marketing-council "is this hook good: <pasted copy>"
47
49
  ```
48
50
 
49
- `--web` opt-in lets the SEO and social-trend seats fetch fresh data. Default runs entirely from training-data expertise (~30 sec, ~$0.10).
50
-
51
- ## Ethical floor
51
+ ## When to use
52
52
 
53
- The skill enforces: no false scarcity, no fake testimonials, no manufactured urgency, no bait-and-switch. Bold claims allowed only when proof-backed.
53
+ Asset-level critique only — review of something that already exists.
54
54
 
55
55
  ## When NOT to use
56
56
 
57
- This skill is built for digital-first, 30-day-horizon, indie-launch marketing. It is NOT built for:
57
+ - **Pre-asset strategy** (which offer first, which channel to enter, 30-day plan before any page exists). The prior 12-seat version mixed strategy and critique, which produced filler. Asset critique and pre-asset strategy belong in separate skills.
58
+ - Multi-decade brand-building, Coca-Cola-tier mass-distribution, SpaceX-tier founder-mission.
59
+ - Regulated industries (medical / financial / legal advice — no FTC / compliance check).
60
+ - Harmful or fraudulent products.
61
+ - Pure ideation with no asset in hand → use `forge-council`.
62
+ - Pure validation of a multi-option decision → use `council-of-12`.
63
+
64
+ ## Architecture changelog
65
+
66
+ ### v2.0.0 (current) — adversarial loop + Foreman
67
+
68
+ - **DELETED** 12 seats (Copywriter, DR, SEO, Brand, YouTube, TikTok, IG, Sales-Psych, Funnel/CRO, Email, Paid-Ads, PR/Community), 30-day launch sequence, Channel Mix priority, SEO Foundation, email welcome / launch sequence, Tensions log.
69
+ - **REPLACED** Hormozi as 12-seat integrator → Hormozi as judge between 2 named adversaries.
70
+ - **PROMOTED** Skeptic from post-verdict challenger → peer adversary across all 3 rounds.
71
+ - **ADDED** substrate intake (3 questions), cheapest disconfirming test ≤72h ≤$50 per Skeptic round, SHIP / EDIT / KILL triple verdict, asset-existence check, generic-phrase blocklist, hard token budget with failure thresholds, Foreman discipline pass (4 checks).
58
72
 
59
- - Coca-Cola-tier mass-distribution brands (no distribution-as-marketing seat)
60
- - SpaceX-tier founder-led mission brands (no 30-year mission narrative)
61
- - Multi-decade brand-building plays
62
- - Regulated industries (medical / financial / legal — no FTC check)
63
- - Harmful or fraudulent products
73
+ Validation evidence: `v3-test-results.md` (TruePoint pre-launch landing page). v2 vs v2.0.0 measurement: tokens dropped from ~2,500 to ~1,600, flaws caught rose from 3 to 8, hallucinated-asset count dropped from ≥1 to 0.
64
74
 
65
- See SKILL.md `## When NOT to Use This Skill` for the full audit.
75
+ ### v1.1.0 (deprecated, kept as rollback anchor)
66
76
 
67
- ## How it was built
77
+ 12-seat panel + Hormozi + post-hoc Skeptic. Shipped Q2 2026. Self-audit on 2026-05-09 measured filler output, fake-distinct headline angles, and Hormozi rubber-stamping.
78
+
79
+ Pin to v1.1.0 if needed:
80
+ ```bash
81
+ git checkout v1.1.0 -- SKILL.md
82
+ ```
83
+
84
+ ## Ethical floor
68
85
 
69
- 3-stage pipeline: Socratic deep-interview (6 rounds, ambiguity gated at 17%) → ralplan consensus (Planner / Architect / Critic, APPROVE) → autopilot execution. Then audited by Council of 12 against Coca-Cola and SpaceX playbooks; Skeptic seat + 4 council recommendations applied (long-arc narrative, founder-voice + spectacle + crisis-response, buyer-POV check, out-of-scope notice).
86
+ The skill enforces: no false scarcity, no fake testimonials, no manufactured urgency, no bait-and-switch. Bold claims allowed only when proof-backed. Hormozi cuts violations from the edit list before Foreman pass.
70
87
 
71
88
  ## Related skills
72
89
 
73
- - `council-of-12` — universal 12-lens analysis with Solomon coordinator (the audit framework that verified this skill)
74
- - `forge-council` — 12-seat creative ideation with Musk coordinator
90
+ - `forge-council` — 12-seat creative ideation with Musk coordinator (pure idea generation, no asset required)
91
+ - `council-of-12` — universal 12-lens analysis with Solomon coordinator (multi-option validation)
92
+ - `marketing-strategy` *(planned)* — pre-asset strategy: offer selection, channel entry, 30-day plan
75
93
  - [silex](https://github.com/ojesusmp/silex) — per-project timeline journal that survives reboots
76
94
 
77
95
  ## Documentation
78
96
 
79
- - [SKILL.md](./SKILL.md) — canonical reference (all 12 disciplines + Hormozi + Skeptic + execution flow + output template)
97
+ - [SKILL.md](./SKILL.md) — canonical reference (4 roles, execution flow, output template, token budget, success criteria)
98
+ - [v3-test-results.md](./v3-test-results.md) — empirical validation: TruePoint Agents pre-launch page, v2 vs v3 vs v3.1 measurement table
80
99
 
81
100
  ## License
82
101
 
83
- This repository ships without an explicit license (all rights reserved by default). To make it freely reusable, the maintainer can add an MIT or Apache-2.0 LICENSE file.
102
+ MIT — see [LICENSE](./LICENSE).
package/SKILL.md CHANGED
@@ -1,648 +1,318 @@
1
1
  ---
2
2
  name: marketing-council
3
- description: "12-seat marketing/sales/copy council with Alex Hormozi as 13th-seat coordinator and a Skeptic post-verdict challenger. Disciplines: Copywriter, Direct-Response, SEO, Brand, YouTube, TikTok, Instagram, Sales-Psych, Funnel/CRO, Email, Paid-Ads, PR/Community. Produces an integrated playbook with positioning (incl. multi-year story arc), 3 distinct headline angles, channel priority + reason, an offer twist via the value equation, a tensions log, a 30-day launch sequence, founder-voice + spectacle + crisis-response plan, buyer-POV check, and a Skeptic counter-argument with a cheap disconfirming test and a failure signal. No web by default; --web opt-in for fresh keyword and trend lookups. Ethical floor: no false scarcity, no fake testimonials, no manufactured urgency. Out of scope: Coca-Cola-tier mass-distribution, SpaceX-tier founder-mission, multi-decade brand-building, regulated industries, harmful products. Use when the user wants to sell, market, position, or promote a digital product, web page, AI agent, course, or indie service."
4
- aliases: [marketing, mkt-council, hormozi-council, mc12, sell-this]
5
- argument-hint: "<product, offer, page, or thing to sell>"
3
+ description: "Adversarial 3-round marketing review with discipline foreman. Pitcher argues the asset works. Skeptic argues it fails with a cheapest disconfirming test under 72 hours. Hormozi judges round 3 with SHIP / EDIT / KILL. Foreman applies four discipline checks (surfaced assumptions, minimum edits, surgical traceability, verifiable test threshold) and produces an amended edit list. Replaces the prior 12-seat panel architecture with named adversaries to kill diffuse responsibility, filler output, fake-distinct headlines, and recommendations for assets that do not exist. Use when the user wants to sell, market, position, audit, or improve any asset (page, copy, offer, hook, ad, email, post). Output target under 1,700 tokens."
4
+ aliases: [marketing, mkt-council, hormozi-council, mc, sell-this]
5
+ argument-hint: "<asset to review: page URL, pasted copy, offer terms, hook, ad, email>"
6
6
  level: 3
7
7
  ---
8
8
 
9
- # Marketing Council and the Hormozi Seat
9
+ # Marketing Council — Adversarial Loop with Foreman
10
10
 
11
- ## Purpose
11
+ ## Why this version exists
12
12
 
13
- The Marketing Council is a universal sales/marketing analysis framework that forces every product, offer, page, or campaign through 12 distinct marketing disciplines before any pitch is made. Most launches fail not because the product is bad, but because the *positioning is wrong, the offer is weak, the hook is generic, or the channel is mismatched.* The Council guarantees no critical marketing angle is left unexamined.
13
+ The prior version used a 12-seat panel + Hormozi integrator + post-hoc Skeptic. Self-audit on 2026-05-09 measured the panel architecture producing filler output, fake-distinct headline angles, premature CTAs for unbuilt assets, and Hormozi rubber-stamping the integration he himself had built. The current version deletes the panel. Two named adversaries (Pitcher and Skeptic) argue 3 rounds. Hormozi judges only after round 3, with both transcripts in front of him. A Foreman then applies four discipline checks and amends the edit list. No diffuse responsibility. No 12 voices to fill space. No coordinator pre-committed to defending their own integration.
14
14
 
15
- Alex Hormozi sits above the 12 seats as the 13th coordinator. He is not a thirteenth perspective — he is the *integrator and the verdict-giver*. He hears every seat, identifies tensions, applies the **value equation** as a single objective integration rule, and signs the playbook. Hormozi never says "consider testing both." Hormozi decides. He picks the option that raises perceived value or lowers buyer risk, and he refuses dark patterns that would cheapen the offer or burn long-term trust.
16
-
17
- The output is one **integrated playbook** the user can execute starting Day 1: positioning, three distinct headlines, a ranked channel mix with a stated reason, an SEO foundation, a funnel + email plan, a PR/community plan, an offer twist, a 30-day launch sequence, a tensions log, Hormozi's one-paragraph sign-off, and the Skeptic's counter-argument.
18
-
19
- ---
20
-
21
- ## When NOT to Use This Skill
22
-
23
- Honest scope check from a Council of 12 audit. This skill is built for **digital-first, 30-day-horizon, indie-launch marketing**: AI agents, SaaS, courses, info products, web pages, indie services. It will produce a weak playbook for these patterns:
24
-
25
- - **Coca-Cola-tier mass-distribution brands.** Skill has no Distribution-as-marketing seat, no mass-media TV/cinema seat, no sponsorship/partnership seat, no localization layer. Coke's playbook is ~50% physical-distribution and ~30% mass-media; this skill covers neither.
26
- - **SpaceX-tier founder-led mission brands.** Skill's PR/Community seat touches founder-voice but does not own a 30-year mission narrative, B2B-government sales, or transparency-about-public-failures as core mechanics.
27
- - **Multi-decade brand-building plays.** This skill plans 30 days. Brand-builders play 30 years.
28
- - **Regulated industries** (medical, financial, legal advice, supplements). Skill has no FTC/regulatory check; user must run claims through legal counsel.
29
- - **Harmful or fraudulent products.** Skill's ethical floor catches dark MARKETING patterns but does not refuse harmful PRODUCTS. Don't use it to sell scams.
30
-
31
- **When to use it:** indie product launches, AI agent / SaaS positioning, single-page sales copy, SEO foundations for new sites, social-content launch sequences, course or info-product offers.
32
-
33
- **When to use something else:** if your product is the next Coca-Cola or the next SpaceX, you need a different skill (or a real agency). This one will give you a competent indie-launch playbook, not a multi-decade brand strategy.
34
-
35
- ---
36
-
37
- ## Quick Reference Card
38
-
39
- | # | Seat | Core Question | Kills | Domain Group |
40
- |---|------|---------------|-------|--------------|
41
- | 1 | Copywriter | "What words make the reader feel the outcome?" | Bland prose | POSITION |
42
- | 2 | Direct-Response | "What makes the reader act NOW, not later?" | Procrastination | POSITION |
43
- | 3 | SEO Specialist | "How do strangers find this when they search?" | Invisibility | DISCOVERY |
44
- | 4 | Brand Strategist | "What does this stand for in one sentence?" | Forgettability | POSITION |
45
- | 5 | YouTube Producer | "What hook earns the next 8 minutes?" | Scrolled-past long-form | DISCOVERY |
46
- | 6 | TikTok Creator | "What hook earns the next 3 seconds?" | Scrolled-past short-form | DISCOVERY |
47
- | 7 | Instagram Strategist | "What earns a save, a share, or a follow?" | Empty grids | DISCOVERY |
48
- | 8 | Sales Psychologist | "What objection is the buyer silently raising?" | Lost-at-checkout | CONVERT |
49
- | 9 | Funnel / CRO Engineer | "Where does the buyer fall off, and why?" | Leaky funnel | CONVERT |
50
- | 10 | Email Marketer | "What earns the next open and the eventual buy?" | Dead lists | CONVERT |
51
- | 11 | Paid Ads Buyer | "What ad creative + audience produces a profitable CAC?" | Wasted spend | DISCOVERY |
52
- | 12 | PR / Community | "Whose endorsement or audience compounds reach for free?" | Lonely launch | AMPLIFY |
53
- | H | **Hormozi** | *Integrates all 12 seats, applies the value equation, signs the playbook with a named tradeoff* | Mediocre offers | — |
54
- | S | **Skeptic** | *Challenges Hormozi's signed playbook — finds the assumption most likely to fail and the cheapest disconfirming test* | Confirmation bias | META |
55
-
56
- **The 4 Domain Groups:**
57
- - **POSITION** (Seats 1, 2, 4): What the product *is* in the buyer's mind. Words, urgency, brand.
58
- - **DISCOVERY** (Seats 3, 5, 6, 7, 11): How strangers *find* the product. SEO, social, paid.
59
- - **CONVERT** (Seats 8, 9, 10): How a viewer *becomes a buyer*. Psychology, funnel, email.
60
- - **AMPLIFY** (Seat 12): How buyers *bring more buyers*. PR, community, partnerships.
61
-
62
- These groups are an organizing tool for the playbook, not separate execution phases. All 12 seats run in one pass.
63
-
64
- ---
65
-
66
- ## Activation Conditions
67
-
68
- ### Explicit Triggers
69
- These phrases activate the Council directly:
70
- - `/marketing-council`, "marketing council", "hormozi council", "mc12"
71
- - "sell this", "how do I sell", "help me sell", "market this", "promote this"
72
- - "what's my hook", "what's my headline", "write me marketing"
73
- - "launch plan", "sales plan", "go-to-market", "GTM"
74
- - "what should I post", "what should I tweet", "TikTok hook", "YouTube angle"
75
-
76
- ### Auto-Trigger Conditions
77
- The Council may recommend activation when detecting:
78
- - A user asking how to sell a product, page, AI agent, service, course, or offer
79
- - A user asking for headlines, hooks, copy, positioning, or a launch plan
80
- - A user pasting product descriptions or landing-page copy and asking "is this good?"
81
- - A user describing a finished build and asking "now what?"
82
-
83
- When auto-triggering, present the recommendation and wait for user confirmation before running. Never auto-execute without consent.
84
-
85
- ---
86
-
87
- ## The 12 Seats — Compact Definitions
88
-
89
- Each seat is a marketing **discipline**, not a persona. Seats do not have opinions or names beyond the discipline. When reporting seat output, say "From the Copywriter's perspective..." not "The Copywriter believes...".
90
-
91
- ---
92
-
93
- ### Seat 1: Copywriter
94
-
95
- **Domain:** POSITION
96
- **Focus:** Words that make the reader *feel* the outcome before they buy.
97
- **Core Question:** "What words make the reader feel the outcome?"
98
-
99
- **Standard Questions:**
100
- 1. What is the single most painful problem this product removes? (Name it in the buyer's own words, not the seller's vocabulary.)
101
- 2. What is the dream-outcome state after the buyer uses this? (Specific, sensory, near-term.)
102
- 3. What is the strongest one-sentence promise that is also TRUE? (No exaggeration, no weasel words.)
103
- 4. What three different angles could the headline take? (Pain-led, outcome-led, curiosity-led, social-proof-led — pick three.)
104
-
105
- **How to analyze:** Strip the product description down to the buyer's pain and the buyer's dream. Re-write both in the buyer's vocabulary, not the founder's. A copywriter who uses the seller's jargon has already lost. Identify the single best one-sentence promise. Then generate three headline angles that pull on different psychological levers (pain / outcome / curiosity / proof) — never three rephrasings of the same idea.
106
-
107
- **Output:** Three distinct headline candidates, each with a one-line rationale. Plus: the single one-sentence promise the page or pitch should lead with.
108
-
109
- **Tagline:** *This seat kills bland prose.*
110
-
111
- ---
112
-
113
- ### Seat 2: Direct-Response
114
-
115
- **Domain:** POSITION
116
- **Focus:** Conversion psychology and the structure of a pitch that ends in action.
117
- **Core Question:** "What makes the reader act NOW, not later?"
118
-
119
- **Standard Questions:**
120
- 1. What is the one specific action the reader must take by the end of this pitch? (Click, sign up, buy, reply.)
121
- 2. What is the strongest reason to act today instead of next month? (Real urgency only — manufactured urgency is forbidden by the ethical floor.)
122
- 3. What objection will most readers raise silently? (And how is it answered before they ask?)
123
- 4. What is the smallest, lowest-risk first step? (Free trial, free chapter, free audit, money-back guarantee.)
124
-
125
- **How to analyze:** Apply the AIDA / PAS / Four U's lens. Map the pitch to: hook → problem → agitation → solution → proof → offer → CTA. Identify which step is weakest. Direct response is verdict-driven: every paragraph either moves the reader closer to the action or it is dead weight.
126
-
127
- **Output:** The CTA wording, the one real urgency reason, and the one risk-reversal element (e.g., guarantee or free trial).
128
-
129
- **Tagline:** *This seat kills procrastination.*
130
-
131
- ---
132
-
133
- ### Seat 3: SEO Specialist
134
-
135
- **Domain:** DISCOVERY
136
- **Focus:** How strangers find this when they type into a search box.
137
- **Core Question:** "How do strangers find this when they search?"
138
-
139
- **Standard Questions:**
140
- 1. What is the buyer typing into Google when they have this exact problem? (Their words, not yours.)
141
- 2. What is the realistic primary keyword (medium-volume, achievable competition)? And what are 3-5 long-tail variants?
142
- 3. What is the page's title tag (under 60 chars) and meta description (under 155 chars)?
143
- 4. What is the one piece of "search intent" content that earns a click — listicle, comparison, how-to, definition?
144
-
145
- **How to analyze:** Without `--web`, rely on training-data knowledge of the niche, common buyer-intent patterns, and standard SEO heuristics (one primary + 3-5 long-tail, intent-matched format, title within length budget). With `--web`, use WebSearch to validate volume and competition; without it, flag estimates as "directional, not measured."
146
-
147
- **Output:** Primary keyword, 3-5 long-tail variants, title tag, meta description, and one content-format recommendation.
148
-
149
- **Tagline:** *This seat kills invisibility.*
150
-
151
- ---
152
-
153
- ### Seat 4: Brand Strategist
154
-
155
- **Domain:** POSITION
156
- **Focus:** Identity, logo direction, and the one-sentence stand-for.
157
- **Core Question:** "What does this stand for in one sentence?"
158
-
159
- **Standard Questions:**
160
- 1. If the product disappeared tomorrow, what would the buyer miss? (That is the brand's job.)
161
- 2. What is the one-sentence brand promise — distinct from the headline, this is the *durable* promise?
162
- 3. What logo direction matches the brand: wordmark, symbol, or combination? What 2-3 colors and what tone (sharp, friendly, premium, raw)?
163
- 4. What is the brand's *enemy* — the thing it stands against? (A brand without an enemy has nothing to rally around.)
164
- 5. What is the multi-year story this brand tells? (SpaceX = "Mars by 20XX." Coca-Cola = "Share happiness, 1971-forever." Without a long-arc story, the 30-day campaign floats alone and forgets itself by month two.)
165
-
166
- **How to analyze:** Brand is the durable layer beneath the campaign. The headline can change weekly; the brand cannot. Identify one durable promise, one enemy (a status quo or a competitor archetype, not a real-named person), a logo + palette direction, and the multi-year story arc the brand will tell. Keep the logo brief: one line of "wordmark in [font style], [color] on [background], anchored by [shape or symbol]" is enough for a designer to start. The long-arc story should fit in one sentence and survive five years of campaigns.
167
-
168
- **Output:** One-sentence brand promise, named enemy, logo brief (one paragraph), 2-3 brand colors + tone, and the one-sentence multi-year story.
169
-
170
- **Tagline:** *This seat kills forgettability.*
171
-
172
- ---
173
-
174
- ### Seat 5: YouTube Producer
175
-
176
- **Domain:** DISCOVERY
177
- **Focus:** Long-form video — videos longer than 3 minutes that earn watch-time and rank.
178
- **Core Question:** "What hook earns the next 8 minutes?"
179
-
180
- **Standard Questions:**
181
- 1. What is the title of the video that someone would click in a YouTube sidebar? (Curiosity gap or specific outcome — under 70 chars.)
182
- 2. What is the thumbnail concept — face, object, contrast, or numeric? (One sentence.)
183
- 3. What is the first 15 seconds — the hook that prevents the close-tab? (One specific scene.)
184
- 4. What is the watch-time-bait — the promise that someone watches all the way through? (Cliffhanger, transformation, payoff.)
185
-
186
- **How to analyze:** YouTube rewards click-through-rate × watch-time. Optimize for both, never one. The thumbnail and title together are the only thing the algorithm sees; the first 15 seconds are the only thing the human gives. Apply MrBeast-style retention thinking without copying his face style — the principle is "every 30 seconds something changes."
187
-
188
- **Output:** Title, thumbnail concept, 15-second hook, and watch-time payoff.
189
-
190
- **Tagline:** *This seat kills scrolled-past long-form.*
191
-
192
- ---
193
-
194
- ### Seat 6: TikTok Creator
195
-
196
- **Domain:** DISCOVERY
197
- **Focus:** Short vertical video — 9 to 60 seconds, native to TikTok and YouTube Shorts and Instagram Reels.
198
- **Core Question:** "What hook earns the next 3 seconds?"
199
-
200
- **Standard Questions:**
201
- 1. What is the *first frame* — the visual or spoken pattern-interrupt that stops the thumb? (Specific object, contrast, or claim.)
202
- 2. What is the *spoken hook* in the first 2 seconds — under 8 words? (Curiosity, controversy, contrast, claim.)
203
- 3. What is the *payoff or twist* — why does someone watch to the end and re-watch?
204
- 4. What audio direction — trending sound, original voiceover, or silent-with-text? (One pick.)
205
-
206
- **How to analyze:** TikTok-style algorithms reward rewatch and complete-rate, not raw views. Engineer the video so the last frame compels a rewatch. The first 2 seconds carry 80% of the retention work. Use the POV / "this is your sign" / "stop doing X, do Y" structures sparingly — only if they fit naturally.
207
-
208
- **Output:** First-frame visual, 8-word spoken hook, payoff/twist, and audio direction. Add a one-line shot list (3-5 cuts).
209
-
210
- **Tagline:** *This seat kills scrolled-past short-form.*
211
-
212
- ---
213
-
214
- ### Seat 7: Instagram Strategist
215
-
216
- **Domain:** DISCOVERY
217
- **Focus:** The mix of feed posts, reels, carousels, and stories that builds a savable, shareable presence.
218
- **Core Question:** "What earns a save, a share, or a follow?"
219
-
220
- **Standard Questions:**
221
- 1. What is the *single post* that, if pinned to the top of the grid, communicates the entire offer in 7 seconds?
222
- 2. What is the *carousel concept* that earns saves — listicle, framework, before/after, mistake-list?
223
- 3. What is the *reel concept* — same vertical-video logic as TikTok, but tuned for the Instagram audience (slightly more polished, less raw)?
224
- 4. What hashtag strategy — niche-tight (3-5 niche tags) or broad-mix (mix of niche + medium + broad)?
225
-
226
- **How to analyze:** Instagram rewards saves and shares more than likes. A post that earns 100 saves outperforms a post that earns 1000 likes. Optimize for the save: that means listicles, frameworks, before/after, and mistake-lists, not pure inspiration posts. The pinned post is the most valuable real estate on the entire grid; design it as a static-mini-landing-page.
227
-
228
- **Output:** Pinned-post concept, one carousel concept (with slide-by-slide outline), one reel concept, and a hashtag strategy.
229
-
230
- **Tagline:** *This seat kills empty grids.*
231
-
232
- ---
233
-
234
- ### Seat 8: Sales Psychologist
235
-
236
- **Domain:** CONVERT
237
- **Focus:** Cialdini's six principles of influence applied to the specific buyer, plus objection-handling.
238
- **Core Question:** "What objection is the buyer silently raising?"
239
-
240
- **Standard Questions:**
241
- 1. Of the six Cialdini principles (Reciprocity, Commitment/Consistency, Social Proof, Authority, Liking, Scarcity), which two are most powerful for THIS buyer? (Pick two; do not invoke all six.)
242
- 2. What are the top three silent objections the buyer raises? ("It's too expensive," "I don't trust this works," "I don't have time," "What if I'm wrong?")
243
- 3. For each silent objection, what is the *evidence-based* answer? (Real proof — case study, demo, guarantee — not a verbal reassurance.)
244
- 4. What is the *commitment ladder* — the smallest first yes that leads to the bigger yes? (Free PDF → email → demo → trial → paid.)
245
- 5. **Buyer-POV check:** if this exact pitch were marketed to YOU, would you feel respected or pressured? (Read the pitch back from the buyer's chair. If "pressured," the playbook has drifted toward Hormozi-aggression — flag it for the Skeptic.)
246
-
247
- **How to analyze:** Buyers don't object out loud; they leave. The job of this seat is to enumerate the silent objections before they happen and make sure the pitch answers each one with evidence rather than rhetoric. Scarcity is allowed only when the constraint is real (limited inventory, real deadline). Manufactured scarcity violates the ethical floor. The buyer-POV check is the one moment in the playbook where the buyer's voice — not the seller's model of the buyer — gets a vote.
248
-
249
- **Output:** Two leveraged Cialdini principles, three silent objections + evidence-based answers, a commitment ladder, and one sentence on the buyer-POV check (respected or pressured, and why).
250
-
251
- **Tagline:** *This seat kills lost-at-checkout.*
15
+ Validation evidence: `v3-test-results.md` in this repo. v2 vs v3 vs v3.1 measurement table at the bottom of that file.
252
16
 
253
17
  ---
254
18
 
255
- ### Seat 9: Funnel / CRO Engineer
19
+ ## Prime Directive
256
20
 
257
- **Domain:** CONVERT
258
- **Focus:** The structure of the landing page and checkout — every box, every button, every form field.
259
- **Core Question:** "Where does the buyer fall off, and why?"
260
-
261
- **Standard Questions:**
262
- 1. What is the page's *single job*? (One conversion event. Not two.)
263
- 2. What is the *above-the-fold structure* — hero (headline + sub-headline + CTA + visual)?
264
- 3. What are the *three proof blocks* below the fold — testimonial, case study, demo, screenshot, logo wall, numbers?
265
- 4. What is the *one objection-handler section* — FAQ, comparison table, guarantee block?
266
-
267
- **How to analyze:** A good landing page has one job. Two jobs = no jobs. Above-the-fold must answer "what is this, who is it for, why should I care, what do I do" in under 5 seconds of scanning. Below-the-fold proves the claim with three different proof types. Below the proof, one objection handler. Below that, the offer + CTA again. Form fields: every extra field cuts conversion by ~5%; ask for the minimum.
268
-
269
- **Output:** Single conversion goal, above-the-fold layout (one paragraph), three proof blocks (one line each), one objection-handler section, and the form-field list (minimum viable).
270
-
271
- **Tagline:** *This seat kills leaky funnels.*
21
+ **This council succeeds when the operator reads the verdict and knows exactly what to ship, kill, or test next — in under 1,700 tokens of output.** It fails when output exceeds 2,000 tokens, when Pitcher and Skeptic converge politely, when Hormozi's verdict could have been written without reading the transcripts, or when Foreman waves through a verdict whose edits collide, exceed evidence, or rest on uncited claims.
272
22
 
273
23
  ---
274
24
 
275
- ### Seat 10: Email Marketer
276
-
277
- **Domain:** CONVERT
278
- **Focus:** The sequences that earn the next open and the eventual buy.
279
- **Core Question:** "What earns the next open and the eventual buy?"
25
+ ## When to use
280
26
 
281
- **Standard Questions:**
282
- 1. What is the *welcome sequence* (5 emails over 7 days)? (Topic per email — value-first, not pitch-first.)
283
- 2. What is the *launch sequence* (5 emails over 5 days)? (Tease → reveal → proof → urgency → close.)
284
- 3. What is the *nurture cadence* between launches? (Weekly newsletter, value-only, with one soft CTA.)
285
- 4. What is the *segmentation* — who gets which list? (At minimum: subscribed-only, demo-attendees, paying-customers.)
27
+ - User asks to sell, market, position, audit, or improve any asset (page, copy, offer, hook, ad, email, post).
28
+ - User pastes copy and asks "is this good" or "what is wrong with this."
29
+ - User asks for headline / CTA / channel decision on a specific asset.
286
30
 
287
- **How to analyze:** Email is the highest-ROI channel and the most-abused one. The welcome sequence sets the tone: 5 value-first emails build the relationship; pitching too early burns the list. The launch sequence is the only time pitching is the main act, and even then the proof must outweigh the pitch. Subject lines matter more than body text; a 30%+ open rate is the bar.
31
+ ## When NOT to use
288
32
 
289
- **Output:** Welcome-sequence outline (5 emails, one line each), launch-sequence outline (5 emails, one line each), nurture cadence, and segmentation.
33
+ - Multi-decade brand-building, Coca-Cola-tier mass-distribution, SpaceX-tier founder-mission, regulated industries (medical / financial / legal advice), harmful or fraudulent products.
34
+ - Pure ideation with no asset in hand → use **forge-council**.
35
+ - Pure validation of a multi-option decision → use **council-of-12**.
36
+ - Pre-asset strategy (which offer first, which channel to enter, 30-day plan before any page exists) → out of scope here. Asset-level critique only. A separate `marketing-strategy` skill is the right home for greenfield planning.
290
37
 
291
- **Tagline:** *This seat kills dead lists.*
38
+ If asset is missing, ask one question and stop: "Paste the asset (URL, copy, offer terms) you want reviewed."
292
39
 
293
40
  ---
294
41
 
295
- ### Seat 11: Paid Ads Buyer
296
-
297
- **Domain:** DISCOVERY
298
- **Focus:** Meta + Google paid acquisition — the ad creative, audience, and budget that produces a profitable CAC.
299
- **Core Question:** "What ad creative + audience produces a profitable CAC?"
42
+ ## Substrate intake (3 questions, mandatory)
300
43
 
301
- **Standard Questions:**
302
- 1. What is the *one ad concept* that should be tested first? (Hook + offer + visual.)
303
- 2. What is the *target audience* — interest-based, lookalike, or retargeting?
304
- 3. What is the *daily budget* needed to learn within 7 days? (Rule of thumb: 50 conversions / week to exit learning phase; budget = target CPA × 50 / 7.)
305
- 4. What is the *pass/fail metric* — CPA target, ROAS target, or click-through threshold?
44
+ Before any role speaks, intake checks the state of the product:
306
45
 
307
- **How to analyze:** Paid ads are an ROI machine, not a brand-builder. Set a target CPA before spending dollar one. The ad creative does 80% of the work; the audience does 20%. Test one strong creative against one tight audience first. Do not test five creatives at once on day one — you will not learn anything statistically. With `--web`, validate current platform best practices; without, use stable principles only.
46
+ 1. **Is the product / asset live?** (URL, deployed copy, working demo? Y/N)
47
+ 2. **Does the product have customers?** (paying or free pilot — Y/N + count if Y)
48
+ 3. **What proof assets exist?** (testimonials, case studies, demo video / audio, screenshots, metrics — list or "none")
308
49
 
309
- **Output:** One ad concept (hook + offer + visual), target audience, daily budget, and pass/fail metric.
50
+ If operator skips intake or answers "unsure," skill defaults to most-conservative substrate (pre-launch, no customers, no proof) and announces this in the verdict.
310
51
 
311
- **Tagline:** *This seat kills wasted spend.*
52
+ Substrate routes which claims are defensible. Pitcher cannot pitch what proof does not back. Skeptic must attack what substrate exposes.
312
53
 
313
54
  ---
314
55
 
315
- ### Seat 12: PR / Community
56
+ ## The 4 Roles
316
57
 
317
- **Domain:** AMPLIFY
318
- **Focus:** Endorsements, partnerships, and community presence that compound reach without paying per impression.
319
- **Core Question:** "Whose endorsement or audience compounds reach for free?"
58
+ ### Role 1: Pitcher
320
59
 
321
- **Standard Questions:**
322
- 1. Who are the *3-5 specific people or accounts* whose endorsement would move the needle? (Named, real, reachable.)
323
- 2. What is the *outreach hook* — the specific reason this person says yes? (Not "would you share this" — a specific, generous reason.)
324
- 3. What is the *community placement* — Reddit threads, Hacker News, Indie Hackers, Twitter/X, LinkedIn, niche Discords? (Pick the 1-2 communities that match the buyer.)
325
- 4. What is the *launch event* — a specific announcement on a specific day with a specific call-to-action? (Product Hunt launch, Show HN, AMA, livestream.)
326
- 5. **Founder voice:** should the founder be visible, and on which platform with what voice? (SpaceX runs on Musk's X account; many indie SaaS companies run on the founder's personal LinkedIn or X. If the answer is "no founder presence," say so — but say so deliberately.)
327
- 6. **Spectacle:** is there a way to make a launch moment so visually impressive that people film it and share it for free? (SpaceX rocket launches earn billions of views at zero ad spend. For software, spectacle could be a public live-build, a 24-hour challenge, or a viral demo. If no spectacle is possible, say so — but try first.)
328
- 7. **Crisis-response template:** if the launch fails publicly (site goes down, demo breaks, competitor attacks), what is the one-paragraph response the founder posts? (Pre-write it before launch. SpaceX broadcasts its rocket explosions; the response is owning the failure and naming the lesson learned. The same approach works for software.)
60
+ **Identity:** You are the Pitcher. Your job is to argue this asset will work. You have read the substrate. You believe in the offer. You are not a yes-man — you make the strongest defensible case, not the strongest possible case. If the asset has a real flaw, acknowledge it and pitch around it; do not lie about it.
329
61
 
330
- **How to analyze:** PR/community is leverage: one endorsement from the right person beats 100 cold ads. Identify named individuals or accounts (not "influencers" generically). Craft a generous-first outreach: lead with what you can do for them, not what you want from them. Pick 1-2 communities with the right audience density; do not spread thin across 6. Founder visibility, spectacle, and crisis-response are the three lessons from SpaceX-tier marketing — they cost almost nothing to plan and dwarf any paid-ad budget when they work.
62
+ **Per round, produce:**
63
+ - **Claim:** one sentence — what this asset will accomplish (conversion, click, sale, trust, ranking).
64
+ - **Mechanism:** one sentence — why it works psychologically or structurally.
65
+ - **Evidence in substrate:** what existing proof or design choice supports the claim. If "none," say "none — claim rests on category pattern."
66
+ - **Strongest single edit:** one specific change that would make the claim more defensible. Quote the current text and the proposed replacement.
331
67
 
332
- **Output:** 3-5 named outreach targets, the outreach hook, 1-2 community placements, one launch-event concept, founder-voice plan (or explicit "no founder presence"), spectacle plan (or explicit "no spectacle available"), and a pre-written one-paragraph crisis-response template.
68
+ **Hard rules for Pitcher:**
69
+ - Cite substrate, not vibes. If proof does not exist, name that.
70
+ - Maximum 4 bullets per round. ~150 tokens.
71
+ - No marketing-jargon ("leverage," "best-in-class," "game-changing"). Cut on detection.
72
+ - No recommending CTAs / assets that do not exist in substrate.
333
73
 
334
- **Tagline:** *This seat kills the lonely launch.*
74
+ ### Role 2: Skeptic
335
75
 
336
- ---
337
-
338
- ## The 13th Seat — Hormozi Coordinator
339
-
340
- Hormozi is not a 13th perspective. Hormozi is the *integrator and the verdict-giver*. He hears all 12 seats, identifies tensions, applies one objective integration rule, and signs the playbook.
76
+ **Identity:** You are the Skeptic. Your job is to argue this asset will fail. You have read the substrate AND the Pitcher's claim. You attack the load-bearing assumption. You are not a contrarian for sport — you find the flaw most likely to break the asset in the wild, then propose the cheapest test that would prove you right.
341
77
 
342
- ### Hormozi's Integration Rule
78
+ **Per round, produce:**
79
+ - **Counter-claim:** one sentence — what will go wrong, and to whom.
80
+ - **Failure mechanism:** one sentence — why it breaks.
81
+ - **Substrate evidence for failure:** what in the substrate (or in what is missing from substrate) makes failure likely.
82
+ - **Cheapest disconfirming test:** one specific test the operator can run in <72 hours, with a binary or numeric pass / fail threshold.
343
83
 
344
- **"Does this make the offer 4x more valuable than the price?"**
84
+ **Hard rules for Skeptic:**
85
+ - Attack one load-bearing assumption per round. Not five small things.
86
+ - Test must be runnable for <$50 and <72 hours. If it is not, it is not the cheapest test.
87
+ - Failure threshold must be binary or numeric. "It might not work well" is forbidden.
88
+ - Maximum 4 bullets per round. ~150 tokens.
345
89
 
346
- Every recommendation from every seat is filtered through this question. If a seat suggests an addition, Hormozi asks: does this raise perceived value, or does it just add noise? If a seat suggests a tactic, Hormozi asks: does this raise the perceived likelihood of the dream outcome, or does it lower trust?
90
+ ### Role 3: Hormozi (judge, round 3 only)
347
91
 
348
- ### The Value Equation
92
+ **Identity:** You are Hormozi. You read all 3 rounds of Pitcher and all 3 rounds of Skeptic. You apply the value equation:
349
93
 
350
94
  ```
351
- Value = (Dream Outcome × Perceived Likelihood of Achievement)
352
- ─────────────────────────────────────────────────────
353
- (Time Delay × Effort & Sacrifice)
95
+ Value = (Dream Outcome × Perceived Likelihood)
96
+ ─────────────────────────────────────
97
+ (Time Delay × Effort & Sacrifice)
354
98
  ```
355
99
 
356
- To raise value, Hormozi asks four questions of every recommendation:
357
- - Does it raise the **Dream Outcome** (bigger / more desirable result)?
358
- - Does it raise **Perceived Likelihood** (more proof, more guarantees, more credibility)?
359
- - Does it lower **Time Delay** (faster results)?
360
- - Does it lower **Effort & Sacrifice** (less work for the buyer)?
361
-
362
- A recommendation that improves any one of those four is a keep. A recommendation that does not is a cut.
363
-
364
- ### Tension Resolution Rule
100
+ You decide which rounds produced real signal and which produced theater. You ship one verdict.
365
101
 
366
- When two seats disagree (Brand says "premium tone," Direct-Response says "punchy CTAs"; Funnel says "long-form proof," TikTok says "3-second hook"), Hormozi resolves by picking the option that **raises perceived value or lowers buyer risk**. If neither does, pick the option that does both least badly. Never split the difference; that produces the least-good version of both.
102
+ **Verdict structure (mandatory, ~400 tokens max):**
103
+ - **Asset in one sentence:** what it currently is in the buyer's mind.
104
+ - **Strongest Pitcher round:** which round (1, 2, or 3), one sentence why.
105
+ - **Strongest Skeptic round:** which round, one sentence why.
106
+ - **Verdict:** SHIP / EDIT / KILL.
107
+ - SHIP = asset works as-is for current substrate.
108
+ - EDIT = asset works with named edits below.
109
+ - KILL = asset fails; reframe needed.
110
+ - **Required edits (if EDIT):** numbered list, max 5. Each edit: location + current text + replacement + reason.
111
+ - **Skeptic test the operator must run:** restate the cheapest test from the Skeptic round that survived. With deadline.
112
+ - **Highest-leverage move:** one sentence — the smallest action that moves value-equation most.
367
113
 
368
- ### Ethical Floor (Non-Negotiable)
114
+ **Hard rules for Hormozi:**
115
+ - Cannot side with whichever role wrote more.
116
+ - Cannot punt ("operator decides"). Must commit verdict.
117
+ - Cannot recommend an asset / CTA that does not exist in substrate. If asset is missing (e.g., "add audio demo"), verdict becomes "EDIT after operator builds X" with X as gating prerequisite.
118
+ - Ethical floor: no false scarcity, no fake testimonials, no manufactured urgency, no bait-and-switch. Cut violations from edit list.
369
119
 
370
- Hormozi enforces these rules on the final playbook before it is shown to the user:
120
+ ### Role 4: Foreman (discipline check, runs after Hormozi)
371
121
 
372
- 1. **No false scarcity.** "Only 3 left" is allowed if there are literally 3 left. "Limited time" is allowed if there is a real deadline. Otherwise, cut.
373
- 2. **No fake testimonials.** Every quote, every case study, every number must be real or marked clearly as illustrative example.
374
- 3. **No manufactured urgency.** Countdown timers are allowed only when tied to a real deadline. Otherwise, cut.
375
- 4. **No bait-and-switch.** The headline must match what the buyer gets. If the offer is paid, do not call the headline "free."
376
- 5. **Bold claims are allowed only when proof-backed.** "10x your output" is allowed only with a case study or data behind it. Without proof, cut to a smaller, defensible claim.
122
+ **Identity:** You are the Foreman. You read Hormozi's signed verdict and apply four discipline checks. You do not re-judge the asset. You audit the verdict for sloppy reasoning, redundant edits, scope creep beyond evidence, and unverifiable thresholds. You produce an amended final edit list. You add ~200 tokens for the right to ship a verdict the operator can act on without rework.
377
123
 
378
- If any seat output violates the ethical floor, Hormozi rewrites it before integration. If a violation cannot be rewritten cleanly, cut the recommendation entirely.
124
+ **The four checks (mandatory, in order):**
379
125
 
380
- ### Verdict Format
126
+ 1. **Surfaced assumptions.** Read every Pitcher and Hormozi claim. Any claim that rests on training-data pattern rather than substrate evidence must be cited or downgraded. Output: list of un-cited claims, each marked `cite or downgrade`.
381
127
 
382
- Hormozi's sign-off is one paragraph at the end of the playbook with three elements:
383
- 1. **The offer in one sentence** — the entire pitch compressed to one line.
384
- 2. **The 4x-value verdict** — does this offer satisfy the value equation? Confidence: HIGH / MEDIUM / LOW.
385
- 3. **The single highest-leverage move** — if the user does only ONE thing tomorrow, what is it?
128
+ 2. **Minimum edits.** Read Hormozi's edit list. Any two edits that touch the same surface (hero CTA area, same paragraph, same form field) must be sequenced or merged. Output: list of edit collisions with proposed sequencing or merge.
386
129
 
387
- ---
388
-
389
- ## Seat S — Skeptic / Devil's Advocate
390
-
391
- The Skeptic is not a 14th discipline. The Skeptic is the **post-verdict challenger**. Hormozi resolves tensions and signs the playbook; the Skeptic then challenges Hormozi's signed playbook from the perspective of someone who has watched 100 launches like this fail. The Skeptic exists because Hormozi (or any single coordinator) carries confirmation bias — once a coordinator has integrated 12 seats, they are emotionally invested in their own synthesis. The Skeptic has no such investment. The Skeptic's job is to find the one thing most likely to break.
130
+ 3. **Surgical traceability.** Read each edit's scope. If an edit says "rewrite all X" but the substrate evidence cited only one specific instance of X, narrow the edit. Output: list of over-broad edits, each with proposed narrower scope tied to the cited evidence.
392
131
 
393
- **Domain:** META (runs AFTER Hormozi signs the playbook)
394
- **Core Question:** "What's the strongest reason this won't work?"
132
+ 4. **Verifiable test threshold.** Read the Skeptic test Hormozi forwarded. The test passes if it has: a binary or numeric pass / fail threshold, a deadline within 72 hours of verdict date, a cost under $50. Output: PASS, or list of missing elements.
395
133
 
396
- **Standard Questions:**
397
- 1. Which assumption in this playbook is most likely wrong, and what evidence would prove it wrong in week 1? (Pick ONE — the load-bearing assumption.)
398
- 2. What is the cheapest disconfirming test the user could run BEFORE spending real money on execution? (Examples: a 5-tweet test before a paid-ad campaign, a landing page with email capture before building the product, a $50 ad spend before a $5,000 launch.)
399
- 3. What is the FAILURE signal — the specific observable that means "stop, this approach is dead, pivot now"? (Not vague. A number, a date, or a binary outcome. Example: "If week-1 click-through is below 1.5% on the cheapest test, stop and re-think positioning.")
400
- 4. Of the 12 seats in the playbook, which one is currently OVERWEIGHTED — i.e., the playbook leans too hard on its recommendation? What happens if that seat is wrong?
401
- 5. Future-trend check: is there a known platform / algorithm / regulatory change in the next 12 months that would invalidate this playbook? (Examples: TikTok ban risk, Apple privacy changes, FTC rule changes, new AI-disclosure laws. If yes, name it and name the contingency.)
134
+ **Foreman output format (~200 tokens):**
135
+ ```
136
+ ## Foreman Pass
402
137
 
403
- **How to analyze:** Disagree on principle, even when the playbook looks strong. The Skeptic's value is exactly proportional to how uncomfortable the counter-argument is for the user to hear. If the Skeptic produces "looks good to me," the Skeptic has failed. The Skeptic does not need to be right — the Skeptic needs to surface the one thing the user would otherwise miss.
138
+ Check 1 — Surfaced assumptions: <PASS, or list of claims to cite-or-downgrade>
139
+ Check 2 — Minimum edits: <PASS, or list of edit collisions + sequencing>
140
+ Check 3 — Surgical: <PASS, or list of over-broad edits + narrowed scope>
141
+ Check 4 — Verifiable test: <PASS, or list of missing elements>
404
142
 
405
- **The Skeptic also enforces the "ask if in doubt" protocol.** If a key fact in the playbook depends on something the user didn't specify (e.g., "I assumed your audience is technical — is that right?"), the Skeptic raises the question explicitly rather than letting the assumption pass.
143
+ ## Foreman-Amended Final Edit List
144
+ 1. <action verb> <edit> — <one-line reason or status: KEEP / NARROWED / SEQUENCED / DEFERRED>
145
+ 2. ...
406
146
 
407
- **Output:** One uncomfortable counter-argument · one cheap pre-launch disconfirming test · one explicit failure signal with a numeric or binary threshold · one overweighted-seat callout · one future-trend risk (or "no known disruptive trends in the next 12 months") · zero or more clarification questions for the user.
147
+ Skeptic test (unchanged or amended): <test + threshold + deadline>
148
+ Highest-leverage move (unchanged or amended): <one sentence>
149
+ ```
408
150
 
409
- **Tagline:** *This seat kills confirmation bias.*
151
+ **Hard rules for Foreman:**
152
+ - Cannot add new edits beyond Hormozi's list. Can keep, narrow, sequence, defer, or merge.
153
+ - Cannot rewrite Hormozi's verdict (SHIP / EDIT / KILL stands).
154
+ - Cannot exceed ~200 tokens. If checks pass, the Foreman section is 4 lines: PASS PASS PASS PASS + restated verdict.
155
+ - If any check finds an issue, name the issue specifically. "Could be tighter" is forbidden. Quote the offending text or list the colliding edits by number.
410
156
 
411
157
  ---
412
158
 
413
159
  ## Execution Flow
414
160
 
415
- Strict sequence. All steps run inside one Claude turn — no subagents, no parallel fan-out, no Read-tool calls during execution.
416
-
417
- 1. **Read the product.** Parse `$ARGUMENTS` for the product / offer / page being marketed. If the description is too thin to analyze (under 10 words and no concrete anchors), ask one clarifying question and stop. Do not invent details.
418
-
419
- 2. **Run all 12 seats in order.** For each seat: state the seat name, ask the standard questions internally, produce the output described under "Output:" for that seat. Keep each seat's output tight — 4-8 lines, not pages. Total seat output should fit in one screen of reading.
420
-
421
- 3. **Hormozi gathers all 12 outputs.** No new analysis at this stage; just collect.
422
-
423
- 4. **Hormozi identifies tensions.** Find every place two seats pull in opposite directions. Common tensions to look for:
424
- - Brand (premium tone) vs Direct-Response (punchy / urgent)
425
- - Funnel (long-form proof) vs TikTok (3-second attention)
426
- - SEO (keyword-stuffed) vs Copywriter (natural voice)
427
- - Sales-Psych (urgency) vs Ethical Floor (no manufactured urgency)
428
- - Paid Ads (cold-traffic angles) vs PR/Community (warm-relationship angles)
429
- At least one tension MUST be named. If none surface, the analysis was too shallow — re-examine.
430
-
431
- 5. **Hormozi resolves each tension** via the value equation: pick the option that raises perceived value or lowers buyer risk. Document the resolution in the tensions log.
432
-
433
- 6. **Hormozi writes the integrated playbook** using the Output Template below.
434
-
435
- 7. **Hormozi appends the 30-day launch sequence** — week-by-week, what gets done, by whom (the user), and what success looks like.
436
-
437
- 8. **Final ethics check.** Scan the entire output for false scarcity, fake testimonials, manufactured urgency, and bait-and-switch. Cut or rewrite any violation. The playbook does NOT ship with violations present.
438
-
439
- 9. **Hormozi signs off.** One paragraph at the end with offer-in-one-sentence + 4x-value verdict + single-highest-leverage-move.
161
+ ```
162
+ 1. Intake (3 substrate questions) → operator answers OR skill assumes most-conservative.
163
+ 2. Round 1: Pitcher claim → Skeptic counter.
164
+ 3. Round 2: Pitcher (revised, addressing Skeptic R1) → Skeptic (new attack vector).
165
+ 4. Round 3: Pitcher (final) → Skeptic (final, must restate strongest test).
166
+ 5. Hormozi reads all 6 outputs → ships raw verdict.
167
+ 6. Foreman runs 4 checks → ships amended final edit list.
168
+ ```
440
169
 
441
- 10. **Skeptic challenges the signed playbook.** Now that Hormozi has signed, the Skeptic reads the entire playbook and produces the counter-argument: load-bearing assumption, cheap disconfirming test, failure signal, overweighted seat, future-trend risk, and any clarification questions the user must answer before executing. If the Skeptic surfaces clarification questions, the playbook ends with those questions explicit — the user must answer them (or accept the assumptions) before Day 1.
170
+ Total: 6 role outputs (~150 each = 900) + Hormozi verdict (~400) + Foreman pass (~200) + intake echo (~100) = **~1,600 tokens target**, hard cap 1,700.
442
171
 
443
172
  ---
444
173
 
445
174
  ## Output Template
446
175
 
447
- Use this exact markdown structure for the playbook:
448
-
449
176
  ```
450
- # MARKETING COUNCIL — <Product Name>
451
- Coordinator: Hormozi | Web: <on|off> | <YYYY-MM-DD>
452
-
453
- ---
177
+ # MARKETING COUNCIL — <Asset Name>
178
+ Date: <YYYY-MM-DD>
454
179
 
455
- ## 1. Positioning
456
- (Brand Strategist + Sales Psychologist)
457
-
458
- **Brand promise (one sentence):** ...
459
- **Brand enemy:** ...
460
- **Two leveraged Cialdini principles:** ... and ...
180
+ ## Substrate
181
+ - Live: Y/N (URL or "not deployed")
182
+ - Customers: Y/N (count)
183
+ - Proof assets: <list or "none">
184
+ - Defaults applied: <list any "unsure" answered as conservative>
461
185
 
462
186
  ---
463
187
 
464
- ## 2. Hook & Headlines
465
- (Copywriter + Direct-Response)
188
+ ## Round 1
189
+ **Pitcher:**
190
+ - Claim: ...
191
+ - Mechanism: ...
192
+ - Evidence in substrate: ...
193
+ - Strongest edit: ...
466
194
 
467
- **Lead promise:** ...
195
+ **Skeptic:**
196
+ - Counter-claim: ...
197
+ - Failure mechanism: ...
198
+ - Substrate evidence for failure: ...
199
+ - Cheapest disconfirming test: ... (threshold: ...)
468
200
 
469
- **Three distinct headline angles:**
470
- - Angle A — pain-led: "..." (rationale)
471
- - Angle B — outcome-led: "..." (rationale)
472
- - Angle C — curiosity- or proof-led: "..." (rationale)
201
+ ## Round 2
202
+ **Pitcher (responding to Skeptic R1):** ...
203
+ **Skeptic (new attack):** ...
473
204
 
474
- **CTA wording:** ...
475
- **Real urgency reason:** ... (or "no real urgency exists; use evergreen CTA")
476
- **Risk reversal:** ...
205
+ ## Round 3
206
+ **Pitcher (final):** ...
207
+ **Skeptic (final, restate strongest test):** ...
477
208
 
478
209
  ---
479
210
 
480
- ## 3. Channel Mix
481
- (YouTube + TikTok + Instagram + Paid Ads)
482
-
483
- **Priority order with reason:**
484
- 1. <channel> — because <audience-match reason>
485
- 2. <channel> — because <reason>
486
- 3. <channel> — because <reason>
211
+ ## Hormozi Verdict (raw)
487
212
 
488
- **YouTube:** title / thumbnail / 15-sec hook / watch-time payoff
489
- **TikTok:** first frame / 8-word hook / payoff / audio
490
- **Instagram:** pinned post / carousel / reel / hashtag strategy
491
- **Paid Ads:** one ad concept / audience / daily budget / pass-fail metric
213
+ **Asset in one sentence:** ...
214
+ **Strongest Pitcher round:** R<n> — ...
215
+ **Strongest Skeptic round:** R<n> — ...
216
+ **Verdict:** SHIP / EDIT / KILL
492
217
 
493
- ---
494
-
495
- ## 4. SEO Foundation
496
- (SEO Specialist)
497
-
498
- **Primary keyword:** ...
499
- **Long-tail variants (3-5):** ..., ..., ..., ..., ...
500
- **Title tag:** ... (under 60 chars)
501
- **Meta description:** ... (under 155 chars)
502
- **Content format:** ...
503
-
504
- ---
505
-
506
- ## 5. Funnel + Email
507
- (Funnel/CRO + Email Marketer)
508
-
509
- **Single conversion goal:** ...
510
- **Above-the-fold:** ...
511
- **Three proof blocks:** ..., ..., ...
512
- **Objection-handler section:** ...
513
- **Form-field list (minimum):** ...
514
-
515
- **Welcome sequence (5 emails):**
516
- 1. ...
218
+ **Required edits (if EDIT):**
219
+ 1. Location: ... | Current: "..." | Replacement: "..." | Reason: ...
517
220
  2. ...
518
- 3. ...
519
- 4. ...
520
- 5. ...
521
221
 
522
- **Launch sequence (5 emails):**
523
- 1. Tease — ...
524
- 2. Reveal — ...
525
- 3. Proof — ...
526
- 4. Urgency — ...
527
- 5. Close — ...
222
+ **Skeptic test the operator must run:** ... | Deadline: <date>
223
+ **Highest-leverage move:** ...
528
224
 
529
- **Nurture cadence:** ...
530
- **Segmentation:** ...
531
-
532
- ---
533
-
534
- ## 6. PR / Community Plan
535
- (PR / Community)
536
-
537
- **3-5 named outreach targets:** ..., ..., ..., ..., ...
538
- **Outreach hook (the generous-first reason):** ...
539
- **1-2 community placements:** ...
540
- **Launch event:** ...
541
-
542
- ---
543
-
544
- ## 7. Offer Twist
545
- (Hormozi value equation applied)
546
-
547
- **Current offer:** ...
548
- **Twist that raises Dream Outcome:** ...
549
- **Twist that raises Perceived Likelihood:** ...
550
- **Twist that lowers Time Delay:** ...
551
- **Twist that lowers Effort & Sacrifice:** ...
552
- **Recommended twist (the strongest of the four):** ...
225
+ — Hormozi
553
226
 
554
227
  ---
555
228
 
556
- ## 8. 30-Day Launch Sequence
557
-
558
- **Week 1 — Foundation:**
559
- - Day 1-2: ...
560
- - Day 3-5: ...
561
- - Day 6-7: ...
562
-
563
- **Week 2 — Content:**
564
- - ...
229
+ ## Foreman Pass
565
230
 
566
- **Week 3 — Launch:**
567
- - ...
231
+ Check 1 — Surfaced assumptions: <PASS, or list>
232
+ Check 2 — Minimum edits: <PASS, or list>
233
+ Check 3 — Surgical: <PASS, or list>
234
+ Check 4 — Verifiable test: <PASS, or list>
568
235
 
569
- **Week 4 — Amplify + Iterate:**
570
- - ...
571
-
572
- ---
236
+ ## Foreman-Amended Final Edit List
237
+ 1. <KEEP / NARROWED / SEQUENCED / DEFERRED> — <edit> — <reason>
238
+ 2. ...
573
239
 
574
- ## Tensions Log
240
+ **Skeptic test:** ... | Deadline: ...
241
+ **Highest-leverage move:** ...
575
242
 
576
- - **<Seat A> vs <Seat B>:** <conflict in one sentence>
577
- - Hormozi resolution: <chosen option> — because <value-equation reason>
578
- - **<Seat C> vs <Seat D>:** <conflict>
579
- - Hormozi resolution: <chosen option> — because <reason>
580
- (Minimum: one named tension. More if they exist.)
243
+ — Foreman
244
+ ```
581
245
 
582
246
  ---
583
247
 
584
- ## Hormozi's Sign-Off
585
-
586
- **The offer in one sentence:** ...
248
+ ## Token Budget + Success Criteria
587
249
 
588
- **The 4x-value verdict:** <HIGH / MEDIUM / LOW> — <one-sentence justification tied to the value equation>
250
+ | Metric | Target | Failure threshold |
251
+ |---|---|---|
252
+ | Total output tokens | ≤1,700 | >2,000 = abort and re-run with tighter constraint |
253
+ | Pitcher / Skeptic per-round tokens | ~150 | >250 = filler |
254
+ | Hormozi verdict tokens | ~400 | >600 = ceremony |
255
+ | Foreman pass tokens | ~200 | >300 = re-judging instead of auditing |
256
+ | Distinct attack vectors across 3 Skeptic rounds | ≥2 | 1 = Skeptic looped |
257
+ | Edits in EDIT verdict (after Foreman) | ≤5 | >5 = unfocused |
258
+ | Generic phrases ("leverage," "best-in-class," "robust," "scalable") | 0 | ≥1 = rewrite |
259
+ | Recommended assets that do not exist in substrate | 0 | ≥1 = critical fail |
589
260
 
590
- **The single highest-leverage move (do this first):** ...
591
-
592
- — Hormozi
261
+ **Self-test before shipping verdict:**
262
+ 1. Did Pitcher and Skeptic disagree in every round? If they converged in any round, re-run that round with sharper opposition.
263
+ 2. Could Hormozi's verdict have been written without reading the rounds? If yes, verdict is theater — rewrite citing specific round content.
264
+ 3. Does the verdict commit to SHIP / EDIT / KILL? If hedged, fail.
265
+ 4. Did Foreman PASS all four checks, or did Foreman name specific issues? If Foreman vague-passed without naming anything, the Foreman section is theater — re-run with stricter audit.
593
266
 
594
267
  ---
595
268
 
596
- ## Skeptic's Counter-Argument
597
- (Runs AFTER Hormozi signs. The Skeptic challenges the signed playbook to kill confirmation bias.)
269
+ ## What this version deletes from v2
598
270
 
599
- **Load-bearing assumption most likely wrong:** ...
600
- **Why it might be wrong:** ...
271
+ | v2 element | v3 disposition | Reason |
272
+ |---|---|---|
273
+ | 12 seats (Copywriter, DR, SEO, Brand, YouTube, TikTok, IG, Sales-Psych, Funnel/CRO, Email, Paid-Ads, PR/Community) | DELETED | Self-audit proved they produced filler + fake variety |
274
+ | Hormozi as integrator-of-12 | REPLACED | Now judge between two adversaries, not synthesizer of twelve |
275
+ | Skeptic as post-verdict challenger | PROMOTED | Now peer adversary across all 3 rounds |
276
+ | 30-day launch sequence | DELETED | Generates wishful timelines; operator's calendar not in scope |
277
+ | SEO Foundation section | DELETED | Was filler in self-audit; not relevant to most substrates |
278
+ | Channel Mix priority | DELETED | Not asset-level; belongs in `marketing-strategy` skill (separate) |
279
+ | Email welcome / launch sequence | DELETED | Coupling failure (recommended sequences for unbuilt assets) |
280
+ | Tensions log | REPLACED | Adversarial structure makes tensions visible inline |
281
+ | Ethical floor | KEPT | Hormozi enforces during verdict |
282
+ | When-NOT-to-use | KEPT (compressed) | Necessary scope guard |
601
283
 
602
- **Cheapest disconfirming test (run this BEFORE spending real money):** ...
603
- **Cost / time of the test:** ...
284
+ ## What this version adds
604
285
 
605
- **Failure signal (stop-and-pivot threshold):** ... (numeric or binary, no vague phrases)
606
-
607
- **Overweighted seat:** <seat name> — what happens if it's wrong: ...
608
-
609
- **Future-trend risk (next 12 months):** ... (or "no known disruptive trend") — contingency: ...
610
-
611
- **Clarification questions for the user (must answer before Day 1):**
612
- - ... ?
613
- - ... ?
614
- (Or "no clarifications needed; assumptions held.")
615
-
616
- — Skeptic
617
- ```
286
+ - **Substrate intake** (3 questions) → forces product-state awareness Pitcher cannot ignore.
287
+ - **Cheapest test ≤72h ≤$50** → every Skeptic round produces a runnable empirical check.
288
+ - **SHIP / EDIT / KILL** triple verdict → forces commitment, no hedging.
289
+ - **Foreman discipline pass** → 4 checks (surfaced assumptions, minimum edits, surgical traceability, verifiable test) catch verdict-level slop Hormozi alone misses.
290
+ - **Token-budget table with failure thresholds** → measurable pass / fail per run.
291
+ - **Generic-phrase blocklist** → catches mush before shipping.
292
+ - **Asset-existence check** → Hormozi cannot recommend assets that do not exist.
618
293
 
619
294
  ---
620
295
 
621
- ## The `--web` Flag
296
+ ## Calibration: how to test this version against v2
622
297
 
623
- By default, the skill runs entirely from training-data expertise. Cost target: ~$0.10, ~30 seconds.
298
+ Run both on the same substrate. Measure:
624
299
 
625
- When `--web` is set in the invocation (e.g., `/marketing-council --web "sell my AI agent to coders"`), specific seats may use web tools:
300
+ 1. **Token count** — current version should be ~60-70% of v2.
301
+ 2. **Flaw-catch rate** — give both a known-flawed asset (e.g., the TruePoint pre-launch page from the 2026-05-09 audit). Count flaws each catches. Current version should match or exceed v2.
302
+ 3. **Operator action clarity** — after reading verdict, operator should be able to list next 3 actions in priority order. If current-version reader can do this faster than v2 reader, current version wins.
303
+ 4. **Hallucinated-asset count** — count recommendations that reference assets not in substrate. Current target: 0. v2 baseline: ≥1 per audit.
626
304
 
627
- - **SEO Specialist** may use **WebSearch** to validate keyword volume and competition.
628
- - **TikTok Creator**, **Instagram Strategist**, and **YouTube Producer** may use **WebFetch** to check current trending audio, hashtags, and topic patterns.
629
- - **Paid Ads Buyer** may use **WebSearch** to check current platform best practices.
630
-
631
- All other seats stay offline regardless of `--web`. This keeps cost predictable: `--web` adds ~$0.30 and ~60 seconds, no more.
632
-
633
- If a seat needs web access but `--web` is not set, the seat must produce its output from training-data knowledge AND mark its findings as "directional, not measured."
305
+ If current version fails any of (1), (3), or (4), revert to v2 via `git checkout v1.1.0 -- SKILL.md`. If only (2) fails, iterate Pitcher / Skeptic prompts.
634
306
 
635
307
  ---
636
308
 
637
- ## Karpathy Compliance Notes
638
-
639
- This skill was scoped under the Karpathy Guidelines. Specifically:
309
+ ## Karpathy compliance notes
640
310
 
641
- - **Single file.** No phase files. No `templates/` directory. No submodule split. council-of-12 has 4 phase files because it has Mode A/B/C; this skill has one mode, so one file.
642
- - **No `--deep` flag.** Speculative. v1 ships a single output mode (integrated playbook). If depth tiers are later requested, refactor THEN.
643
- - **No Mode A/B/C.** All 12 seats run every time. Speculative tiering is a non-feature.
644
- - **No automated test harness.** Single benchmark: `/marketing-council "sell my AI agent to coders"` must produce three distinct headlines, channel + reason, offer twist via value equation, tensions log with at least one Brand-vs-DR conflict, and pass the no-slop test.
645
- - **Two named roles: Hormozi (coordinator) and Skeptic (post-verdict challenger).** The 12 numbered seats are *disciplines*, not named marketers. No Ogilvy seat, no Halbert seat, no MrBeast seat — those would be a different skill (forge-marketers or similar). The Skeptic was added after a Council of 12 audit found that any single coordinator carries confirmation bias; the Skeptic exists specifically to challenge Hormozi's signed playbook.
311
+ - **Single file.** No phase files. No `templates/` directory. No submodule split.
312
+ - **No `--deep` flag.** Speculative. v2.0.0 ships a single output mode.
313
+ - **No mode tiers.** All 4 roles run every time. Speculative tiering is a non-feature.
314
+ - **Four named roles only:** Pitcher, Skeptic, Hormozi, Foreman. No Ogilvy seat, no Halbert seat, no MrBeast seat — those would belong to a different skill.
315
+ - **Foreman implements the 4 karpathy principles** (surfaced assumptions, minimum edits, surgical traceability, verifiable success criteria) without naming the framework. Principles only; no name-drop in user output.
646
316
 
647
317
  ---
648
318
 
@@ -650,44 +320,19 @@ This skill was scoped under the Karpathy Guidelines. Specifically:
650
320
 
651
321
  **These rules are non-negotiable. They override any conflicting interpretation. Read them before executing. Read them again after executing.**
652
322
 
653
- ### Identity Rules
654
-
655
- 1. **Seats are DISCIPLINES, not personalities.** Never say "The Copywriter believes..." — say "From the Copywriter's perspective..." or "The copywriting lens reveals..." Seats analyze; they do not opine.
656
-
657
- 2. **Hormozi DECIDES; seats ADVISE.** Hormozi is not a moderator. Hormozi is the verdict-giver. If the playbook ends without a clear sign-off, Hormozi has failed.
658
-
659
- 3. **The Council is not a debate.** Seats do not argue with each other. They each analyze independently. Tensions are identified by Hormozi during integration, not during seat execution.
660
-
661
- ### Output Rules
662
-
663
- 4. **Always produce three DISTINCT headline angles.** Three rephrasings of the same idea is a fail. The three must pull on different psychological levers (pain / outcome / curiosity / proof — pick three).
664
-
665
- 5. **Always state the channel-priority REASON.** "Use TikTok" is not enough. "Use TikTok because indie devs scroll TikTok during context-switches and the algorithm rewards specific tech-tip hooks" is enough.
666
-
667
- 6. **Always include at least one offer twist via the value equation.** Identify one of the four levers (Dream Outcome / Likelihood / Time / Effort) and apply it. "Add a guarantee" without naming the lever is a half-twist.
668
-
669
- 7. **Always include a tensions log with at least one named conflict.** If no tension surfaces, the analysis was shallow — re-examine. The most common real tension is Brand-vs-Direct-Response on tone.
670
-
671
- 8. **Never ship dark patterns.** Hormozi's ethical floor is enforced last, before output. False scarcity, fake testimonials, manufactured urgency, bait-and-switch — cut.
672
-
673
- ### Quality Rules
674
-
675
- 9. **Never use generic phrases.** "Leverage synergies." "Best-in-class." "Revolutionary." "Game-changing." "Take it to the next level." If any of these phrases appears in the output, rewrite that section. Specific, concrete language only.
676
-
677
- 10. **Every recommendation must be EXECUTABLE on Day 1.** "Build a brand" is not executable. "Use a wordmark in a sharp sans-serif (Inter, Geist, or similar), in deep navy on warm-white, with a small lightning-bolt symbol left of the word, on the landing page hero by Friday" is executable.
678
-
679
- 11. **The playbook must be DECIDABLE.** A non-expert user must read it and start executing without re-asking. If a section is vague enough that the user has to come back and ask "but how?", that section failed.
680
-
681
- 12. **Do not manufacture findings.** If a seat genuinely has nothing to add for a particular product (e.g., Paid Ads for a free hobby project with no budget), say "Paid Ads — not relevant for this stage; revisit when budget exists" and move on. Inventing recommendations to fill space is worse than acknowledging the gap.
682
-
683
- 13. **The Skeptic must produce DISCOMFORT.** The Skeptic's value is exactly proportional to how uncomfortable the counter-argument is for the user to hear. "Looks good, no concerns" is a Skeptic failure — rewrite. The Skeptic must always name a load-bearing assumption, a cheap disconfirming test, and a failure signal with a real threshold. If the Skeptic surfaces clarification questions, those questions are part of the playbook output — the user must see them and answer them before Day 1.
684
-
685
- 14. **Out-of-scope products: refuse early, not after the playbook.** If the user invokes the skill on a Coca-Cola-tier mass-distribution brand, a SpaceX-tier founder-led mission brand, a multi-decade brand-building play, a regulated industry needing legal review, or a fraudulent / harmful product — say so at the START of the run, before producing the playbook. Do not produce a weak playbook and then disclaim it; refuse politely and point to the "When NOT to use this skill" section.
686
-
687
- ### The Prime Directive
688
-
689
- **This Council exists to convert vague product ideas into executable marketing playbooks for non-experts. It succeeds when the user reads the output and knows exactly what to do tomorrow morning. It fails when it produces generic AI marketing-speak that anyone could have written without the framework. Every section must make the user think "I would not have thought of that, and I can do this." If it does not, the section is dead weight — cut it.**
323
+ 1. **Roles are ADVERSARIES, not personas.** Pitcher does not "feel" the asset will work; Pitcher argues it will. Skeptic does not "worry" it might fail; Skeptic argues it will. Hormozi does not "consider" — Hormozi judges. Foreman does not "suggest" — Foreman audits.
324
+ 2. **Hormozi DECIDES; Foreman AUDITS; neither moderates.** If the verdict ends without SHIP / EDIT / KILL, Hormozi failed. If Foreman waves through without naming specifics on at least one check, Foreman failed.
325
+ 3. **Always run all 3 rounds.** Skipping rounds = panel re-emergence = filler.
326
+ 4. **Always produce a binary or numeric Skeptic test.** "Try it and see" is forbidden.
327
+ 5. **Never ship dark patterns.** Hormozi's ethical floor is enforced last, before output. False scarcity, fake testimonials, manufactured urgency, bait-and-switch — cut.
328
+ 6. **Never use generic phrases.** "Leverage synergies." "Best-in-class." "Revolutionary." "Game-changing." If any appear in the output, rewrite that section.
329
+ 7. **Every recommendation must be EXECUTABLE on Day 1.** "Build a brand" is not executable. "Replace hero CTA `Get in touch` with `Hear how it works (2-min call)`" is executable.
330
+ 8. **Do not manufacture findings.** If a substrate genuinely lacks proof for a claim, name that. Inventing recommendations to fill space is worse than acknowledging the gap.
331
+ 9. **The Skeptic must produce DISCOMFORT.** "Looks good, no concerns" is a Skeptic failure — rewrite. The Skeptic must always name a load-bearing assumption, a cheap disconfirming test, and a failure signal with a real threshold.
332
+ 10. **Out-of-scope assets: refuse early, not after the verdict.** If the user invokes the skill on a Coca-Cola-tier brand, a regulated industry, or a fraudulent product — say so at the START of the run.
333
+ 11. **Foreman never adds edits.** Keep, narrow, sequence, defer, merge — only.
334
+ 12. **If Foreman finds nothing, output 4 PASS lines and stop.** Do not pad.
690
335
 
691
336
  ---
692
337
 
693
- *Marketing Council. Twelve disciplines. One Hormozi who decides. One Skeptic who challenges. One playbook. No dark patterns, no generic mush, no half-finished pitches, no unchallenged confidence.*
338
+ *Marketing Council v2.0.0. Three rounds. Two adversaries. One Hormozi who decides. One Foreman who audits the decision. No diffuse responsibility, no filler, no fake-distinct headlines, no hallucinated assets, no unverifiable tests.*
package/package.json CHANGED
@@ -1,7 +1,7 @@
1
1
  {
2
2
  "name": "@ojesusmp/marketing-council",
3
- "version": "1.1.0",
4
- "description": "12-seat marketing/sales/copy council with Hormozi coordinator and Skeptic post-verdict challenger. Claude Code skill for selling digital products, AI agents, web pages, courses, and indie services.",
3
+ "version": "2.0.0",
4
+ "description": "Adversarial 3-round marketing review with discipline foreman. Pitcher / Skeptic / Hormozi / Foreman. Claude Code skill for asset-level critique of pages, copy, offers, hooks, ads, and emails. Output under 1,700 tokens.",
5
5
  "bin": {
6
6
  "marketing-council": "install.js"
7
7
  },
@@ -17,9 +17,9 @@
17
17
  "marketing",
18
18
  "sales",
19
19
  "copywriting",
20
- "seo",
21
20
  "hormozi",
22
- "council-of-12",
21
+ "adversarial-loop",
22
+ "asset-critique",
23
23
  "skill",
24
24
  "anthropic"
25
25
  ],