A 6-layer AI skill chain powered by an enterprise-grade deep research engine that produces consulting-grade Brand Strategy & Competitive Positioning Reports.
Built by Branded Mayhem Collective for Claude. Works in Claude.ai (web), Claude Code, Claude desktop, Cursor, and Codex CLI — anywhere SKILL.md is read.
The Brandprint Engine uses an 8-phase deep research pipeline (scope, plan, retrieve, triangulate, synthesize, critique, refine, package) to gather evidence from hundreds of sources, then chains that evidence through 6 brand strategy layers to produce a 25-35 page strategic deliverable — the same kind of report that Strategy&, EY-Parthenon, Monitor Deloitte, and KPMG Strategy charge $50K-200K to produce.
Input: A brand name, audience description, and 3-5 competitor URLs.
Output: A complete Brand Strategy & Competitive Positioning Report with:
- Core Human Truth (the foundational buyer insight no one else articulates)
- 10 evidence-backed strategy elements (audience segments, JTBD, equity ladder, economic engine)
- Brand Activator, Platform Word, Tone & Manner, Moat Statement, and 10 Brand Guardrails
- Brand Mantra (3 words), Personification Vignette, Header Translation Line
- Competitive white space analysis with collision scoring
- Implementation roadmap (4 phases, 24 months)
- Full source bibliography (typically 400-1000+ research sources)
- Professional HTML & PDF output (McKinsey-style formatting)
| Command | Purpose |
|---|---|
/brandprint-engine:intake "{slug}" |
Start here — intake interview → inputs.yaml + state |
/brandprint-engine:run --engagement {slug} |
Layers 1-4, subagent per layer, checkpointed + schema-validated |
/brandprint-engine:audit --engagement {slug} [--auto] |
Layer 5 competitive gate — PROCEED/REFINE |
/brandprint-engine:compile --engagement {slug} |
Final 25-35pp report (markdown + PDF) |
/brandprint-engine:status |
Where every engagement stands |
/brandprint-engine:doctor |
Diagnose the installation |
/brandprint-engine:configure |
Firm branding + research/report defaults |
/brandprint-engine:feedback |
Capture learnings — every future run gets sharper |
Engagement state lives in .brandprint/engagements/{slug}/. See CHANGELOG.md
for what v2 adds over the 1.x manual chain (which still works and remains the
Claude.ai web path).
┌─────────────────────────────────────────────────────┐
│ DEEP RESEARCH ENGINE │
│ 8-phase pipeline: scope → plan → retrieve → │
│ triangulate → synthesize → critique → refine → │
│ package │
│ │
│ • 4 research modes (Quick/Standard/Deep/UltraDeep) │
│ • Parallel web search (5-10 concurrent queries) │
│ • Source credibility scoring (0-100) │
│ • Citation verification + anti-hallucination │
│ • Progressive file assembly for unlimited length │
└──────────────┬──────────────────────────────────────┘
│ powers every research-heavy layer
▼
Layer 1: Core Human Truth ─── one validated sentence capturing buyer tension
|
v
Layer 2: Brandprint Tier-A ── 10 strategy elements with evidence gates
|
v
Layer 3: Brandprint Tier-B ── 5 actionable elements with behavioral proxy tests
|
v
Layer 4: Brandprint Tier-C ── 4 stylistic elements ready to deploy
|
v
Layer 5: Competitive Audit ── validates differentiation vs. named competitors
|
+-- If REFINE --> re-run Layers 1-4 with competitive context (the magic)
|
+-- If PROCEED --v
|
Layer 6: Report Compiler ─── 25-35 page consulting-grade deliverable
(HTML + PDF via McKinsey template)
The research engine is what separates this from "ask AI to brainstorm a brand strategy." Every research-heavy layer (1, 2, and 5) is powered by an enterprise-grade 8-phase pipeline:
| Phase | What It Does |
|---|---|
| Scope | Decomposes the research question, identifies stakeholder perspectives, defines boundaries |
| Plan | Maps sources, creates search query strategy with variants, plans triangulation |
| Retrieve | Launches 5-10 parallel searches across multiple angles + spawns sub-agents for deep dives |
| Triangulate | Cross-references facts across 3+ independent sources, flags contradictions |
| Outline Refinement | Adapts structure based on what the evidence actually shows (not what we assumed) |
| Synthesize | Identifies patterns, maps relationships, generates insights beyond individual sources |
| Critique | Multi-persona red teaming: Skeptical Practitioner, Adversarial Reviewer, Implementation Engineer |
| Package | Structures output with full bibliography, generates HTML/PDF if needed |
Research modes:
- Quick (3 phases, 2-5 min) — initial exploration
- Standard (6 phases, 5-10 min) — most brand research questions
- Deep (8 phases, 10-20 min) — competitive audits, complex topics
- UltraDeep (8+ phases, 20-45 min) — comprehensive category reviews
Quality controls:
- Every claim cited immediately with source reference
- Source credibility scored 0-100 (domain authority, recency, expertise, bias)
- Citation verification (DOI/URL validation, hallucination detection)
- 9-point automated report validation
This is what makes the Brandprint Engine different from running a single AI prompt.
First pass (Layers 1-4): Builds positioning from the brand's own tensions and evidence. Good output, but untested against competitors.
Layer 5: Stress-tests every claim against named competitors. Scores semantic collision, vocabulary overlap, and territory ownership. Returns PROCEED or REFINE.
Second pass (if REFINE): Re-runs Layers 1-4 with competitive context baked in as constraints. The output is dramatically sharper because the engine now knows which territories are occupied, which claims collide, and where open space exists.
Real example: First pass produced "We build Texas. Our crews own it." — which collided with a competitor's entrenched "Texas' Premier Builder" positioning. Second pass (after competitive audit) produced "Complexity-first builder" + "Same Team, Every Project" + "Certainty" as platform word. Zero competitor overlap. Category of one.
- A Claude account — works in Claude.ai (web), Claude Code (CLI), Claude desktop, Cursor, or Codex CLI. Any surface that reads SKILL.md.
- Claude Max or Pro subscription — the research phases are token-intensive
- Web search — built-in WebSearch works out of the box. For enhanced research, optionally install search-cli (
brew tap 199-biotechnologies/tap && brew install search-cli) - Python 3.9+ — for research engine scripts (validation, citation verification, HTML/PDF generation)
- Optional: weasyprint (
pip install weasyprint) — for PDF output
This repo works on two surfaces — pick yours. The skills (and the deep-research scripts that back them) are surface-aware: paths resolve correctly in both Claude Code and Claude.ai web automatically.
Claude Code (plugin install — primary path):
/plugin marketplace add Branded-Mayhem-Collective-LLC/brandprint-engine
/plugin install brandprint-engine@bmc-brandprint-engine
Installing the whole stack? Add Branded-Mayhem-Collective-LLC/contraband-marketplace and install with @contraband instead.
On enable, a SessionStart hook checks Python 3.9+ and (optionally) weasyprint, and a PostToolUse hook auto-runs validate_report.py on any brandprint deliverable when Claude writes it. Warnings only — nothing blocks.
Claude.ai web / Claude Desktop:
In Claude.ai web, paste the repo URL into the chat and ask:
"Install the skills from https://github.com/Branded-Mayhem-Collective-LLC/brandprint-engine"
Claude fetches, registers the skills, and asks permission. Alternatively download the repo as a .zip (GitHub Code button → Download ZIP), unzip locally, then upload the skill folders via Settings → Customize → Skills. The same SKILL.md files run on both surfaces — paths inside resolve automatically.
In Claude Code:
/plugin list
You should see brandprint-engine enabled, with 10 skills available under the namespace /brandprint-engine:<skill>.
In Claude.ai web: ask Claude "list the brandprint skills you have access to" — you should see all 10.
/plugin update brandprint-engine
Updates flow through this repo's main branch on every version bump in plugin.json.
Run the core-human-truth skill for [BRAND NAME].
Topic: [what they do]
Audience: [who buys from them]
Region: [geographic market]
Mode: standard
The skills auto-chain. If they don't, prompt each tier manually:
- "Run brandprint-tier-a using the Core Human Truth output as seed."
- "Run brandprint-tier-b using the Tier-A output as seed."
- "Run brandprint-tier-c using the Tier-A and Tier-B outputs as seeds."
Run the competitive-positioning-audit for [BRAND NAME].
Competitors:
1. [Competitor 1] - [URL]
2. [Competitor 2] - [URL]
3. [Competitor 3] - [URL]
Pro tip: Paste actual competitor website copy for sharper collision analysis.
Re-run Layers 1-4 with the competitive context from the audit. Then re-run Layer 5 to confirm PROCEED.
Run the brand-strategy-compiler.
Brand: [BRAND NAME]
Client: [CLIENT NAME]
Prepared by: [YOUR FIRM NAME]
You now have a 25-35 page Brand Strategy & Competitive Positioning Report ready for client presentation.
brandprint-engine/
|-- README.md # This file
|-- skills/
|-- deep-research/ # Research engine
| |-- SKILL.md # Entry point (102 lines)
| |-- reference/ # 6 methodology docs
| |-- scripts/ # 7 Python utilities
| +-- templates/ # Report + HTML templates
|-- core-human-truth/
| +-- SKILL.md # Layer 1 (185 lines)
|-- brandprint-tier-a/
| +-- SKILL.md # Layer 2 (299 lines)
|-- brandprint-tier-b/
| +-- SKILL.md # Layer 3 (257 lines)
|-- brandprint-tier-c/
| +-- SKILL.md # Layer 4 (244 lines)
|-- competitive-positioning-audit/
| +-- SKILL.md # Layer 5 (183 lines)
|-- brand-strategy-compiler/
| +-- SKILL.md # Layer 6 (284 lines)
+-- brandprint-engine-guide/
| +-- SKILL.md # Runner guide (316 lines)
|-- competitive-teardown/
| +-- SKILL.md # Advanced playbook (221 lines)
+-- brand-revival/
+-- SKILL.md # Advanced playbook (246 lines)
Total: 10 skill packages including the deep research engine with 7 Python scripts, 6 reference docs, and 2 templates, plus 2 advanced playbooks. v2 adds 8 commands (commands/), 7 layer agents (agents/), JSON output schemas (templates/schemas/), engagement state manager (scripts/brandprint_state.py), and a pytest suite (tests/).
For step-by-step instructions, troubleshooting, tips for best results, and the full two-pass workflow explanation, run:
/brandprint-engine-guide
Or read skills/brandprint-engine-guide/SKILL.md directly.
An 8-phase research pipeline that powers Layers 1, 2, and 5. Launches parallel web searches across multiple angles, spawns sub-agents for specialized deep dives, triangulates findings across 3+ independent sources, and scores source credibility (0-100). Includes citation verification, anti-hallucination protocols, and progressive file assembly for reports of any length. Outputs: verified evidence base with full bibliography.
Uses the deep research engine in Standard or Deep mode to search across academic sources, industry data, forums, and reviews to find the single sentence (max 25 words) that captures the foundational tension buyers feel but rarely articulate. Produces: truth sentence, tension ladder, buyer archetypes, resonance/repellence lexicon.
10 evidence-gated sprints building on the Core Human Truth: audience segments, desired outcomes (JTBD), target audience sizing, competitive set, signature offers, expertise signals, economic engine, core equities, equity ladder, and one-line promise/tagline. Each sprint has independent proof minimums and claim sheets.
5 actionable elements with behavioral proxy tests: Brand Activator (the signature micro-behavior), Brand Platform Word (the single-word decision filter), Tone & Manner (adjectives + key phrases from real customer language), Moat Statement (why you win, with observable signals), and Brand No-No's (10 guardrails with rationale).
4 stylistic elements derived entirely from Layers 2-3 (no new claims): Brand Mantra (exactly 3 words), Personification Vignette (day-in-the-life character), Header Translation Line (max 20 words, SVO structure), and Features List mapped to the Equity Ladder with zero orphans.
Uses the deep research engine in Deep or UltraDeep mode for comprehensive competitor intelligence, then runs 5-phase adversarial analysis: Forensic Collection (exact competitor language), Collision Analysis (0-10 differentiation scoring per claim), Buyer Perception Modeling (side-by-side comprehension testing), Repositioning Paths (3-5 alternatives with evaluation matrix), and Second-Pass Decision (PROCEED or REFINE gate).
Transforms all layer outputs into an 11-section consulting-grade narrative: Executive Summary, Market Context, Core Human Truth & Buyer Psychology, Competitive Landscape & White Space, Target Segments, Brand Architecture & Equity Framework, Competitive Moat, Signature Offers & Growth Strategy, Brand Activation System, Strategic Guardrails, and Implementation Roadmap.
These are standalone strategic playbooks that can be used independently or chained with the core 6-layer engine for deeper analysis.
Category-level disruption analysis using a SCAMPER-across-the-matrix approach. Maps all major players into a feature/value matrix, applies each SCAMPER lens to the category's collective blind spots (not individual brands), identifies structurally empty quadrants, and develops 2-3 concepts that exploit those gaps. Includes feasibility filtering against client capabilities. Feeds into the Competitive Positioning Audit (Layer 5) as a pre-built competitive matrix, or operates as a standalone research sprint.
A 5-phase playbook for resurrecting stale, broken, or dormant brands. Triages the failure mode (Stale/Broken/Invisible), runs a root-cause autopsy, inventories surviving brand assets, designs unexpected cultural partnerships (the "surprising but inevitable" model), structures a 3-act limited drop calendar over 12-18 months, and builds an authenticity bridge with collaborators native to where the brand is going. Outputs a Brand Revival Prescription (Symptoms, Root Cause, Treatment, Dosage, Prognosis). Draws from Core Human Truth (Layer 1) and Competitive Positioning Audit (Layer 5) when used within the engine.
How long does the full chain take? 2-4 hours with the deep research engine in Standard/Deep mode. In Quick mode, 1-2 hours but with lighter evidence. UltraDeep mode can extend Layer 5 to 45+ minutes but produces the most thorough competitive intelligence.
Can I run individual layers standalone? Yes. Each skill works independently. But the chain produces the best results because each layer builds on validated outputs from previous layers.
What if I don't have competitors to test against? Layer 5 requires at least one named competitor. If you truly have no competitors, you can skip to Layer 6 — but the report will note that competitive validation was not performed.
Can I update the Brandprint later? Yes. Save your layer outputs. To update (e.g., after 6 months), re-run Layer 5 with current competitor positioning. If the market has shifted, re-run the full chain.
What industries does this work for? Any. The skills are industry-agnostic. The research adapts to whatever category, audience, and region you specify. It has been tested on construction, SaaS, professional services, food & beverage, and retail.
Is the output actually consulting-grade? The research depth (400-1000+ sources) often exceeds what a human team reviews. The 11-section report structure matches the format used by top strategy firms. The difference is speed and cost, not quality.
MIT. Free and open — use it, fork it, ship client work with it. Outputs you produce with these skills are yours. See LICENSE.
Questions or issues? Contact michael@brandedmayhem.com
Built by Branded Mayhem Collective | Dallas, Texas