Product Marketing Context

Product Marketing Context

The brain every other PMM skill reads before doing anything. Build it once. Every subsequent skill — positioning, competitive, GTM strategy, stakeholder maps — pulls from it automatically. Zero re-entry.

Not a form. A 15-minute conversation that makes every future PMM output sharper.

---

Trigger

• When:
• Building GTM context from scratch (new company, new product, new role)
• Auditing whether your current brain still matches what you're telling customers
• Refreshing buyer intelligence after a market shift or repositioning event
• Any PMM skill is about to run and there's no brain file yet
• Checking brain health and getting improvement recommendations
• Not for: One-time positioning work → hs-positioning-messaging. Competitive

analysis → hs-competitive-battlecard. Individual audience research → hs-buyer-personas. OKR planning → hs-brainstorm-okrs. Each of those skills reads your brain — this skill builds it.

• Example prompts:
• "Build my brain"
• "Set up my GTM foundation"
• "Audit my brain"
• "Check my brain health"
• "Update my ICP"
• "What's my current brain?"
• "Configure my PMM context"

---

Inputs

• Args: None required. Skill auto-detects whether a brain exists or needs building.

Optional: --edit-section [1-6] to update one section without re-running the full wizard.

• Defaults:
• No brain found → first-run onboarding (see Pre-flight) → fresh wizard
• Brain exists → offer: [View current] [Edit sections] [Rebuild from scratch]
• v1.x legacy files found in .agents/ or .claude/ → offer migration with

field-by-field confirmation before writing

• Context keys:
• /foundation/brain.md — read if exists
• .agents/product-marketing-context.md — check for legacy v1.x
• .agents/icp.md — check for legacy v1.x
• .agents/alternatives-map.md — check for legacy v1.x
• .agents/voice-tone.md — check for legacy v1.x
• .agents/market-context.md — check for legacy v1.x
• .agents/proof-points.md — check for legacy v1.x

---

Pre-flight

First-run detection (Ramp Glass pattern)

Before anything else: check for /foundation/brain.md.

If brain is missing AND this is first contact with the skill, run the first-run hook. Do not go straight to wizard questions. Show value before asking for effort:

> "Before we start: what you're about to build is the file every PMM skill in this > ecosystem reads automatically. ICP, positioning, alternatives, voice, proof points — > captured once here, never re-entered. Skills like hs-positioning-messaging, > hs-competitive-battlecard, and go-to-market-strategy will pull from it without > you typing a word. > > It takes 15 minutes. You do it once. Ready to start?" > [Yes, let's build it] [Not now — remind me later] [I have old files to migrate]

If brain exists: skip first-run hook. Load silently, offer [View / Edit / Rebuild].

If legacy v1.x files found (in .agents/ or .claude/): offer migration: > "Found older foundation files from v1.x. Want to migrate them automatically? > I'll show you each extracted value before writing anything." > [Yes, migrate] [No, start fresh]

Standard checks

• /foundation/ directory — create if missing
• Any section marked 🔴 Placeholder → flag before running dependent skills:

> "Section [X] is incomplete. Skills that depend on it will produce weaker output."

---

Steps

Step 1: Detect Brain State

Check what exists and branch accordingly:

• /foundation/brain.md present → offer [View / Edit section / Rebuild]
• Legacy v1.x files in .agents/ → offer migration
• Neither → first-run hook (Pre-flight) → fresh wizard

If editing: ask which section (1–6) or "all". Jump directly to that section's questions without re-running the others.

---

Step 2: Product Context (5 Questions)

Why this section: Every skill needs to know what you're selling and to whom. Without it, positioning output is generic. This anchors everything.

Q1: What's your product called? Q2: In 2–3 sentences, what does it do? Focus on the problem it solves, not features. Q3: What stage is it at? (Pre-launch / Launch / Growth / Scale) Q4: Who's your primary target market? Be specific — company size, industry, role. Q5: Complete: "[Product] helps [target market] to _____ so they can _____"

Store as: product_name, product_description, product_stage, target_market, value_proposition

Confirm Section 1: Show answers. "Look good? (yes / edit / skip for now)"

---

Step 3: ICP Definition (8 Questions)

Why this section: Vague ICP = vague messaging. Every persona, battlecard, and launch brief traces back to this. Specificity here multiplies across every output.

Q1: What company size (employees or ARR) is your ideal customer? Q2: Any specific industries or verticals? Or horizontal across all industries? Q3: Primary geographic market(s)? Q4: Who's the primary buyer? (Title + role description) Q5: Who actually signs the contract? (Economic buyer, if different) Q6: Who becomes your internal champion? What do they care about? Q7: What are the top 3 pain points your ICP has that your product solves? Q8: What event triggers them to start looking for a solution like yours?

Store as: icp_company_size, icp_industry, icp_geography, icp_primary_persona, icp_economic_buyer, icp_champion, icp_pain_points[], icp_buying_triggers

Confirm Section 2: Show answers. "Look good? (yes / edit / skip for now)"

---

Step 4: Alternatives & Positioning (6 Questions)

Why this section: April Dunford's framework. You're not positioning against "competitors" — you're positioning against what buyers are actually comparing you to, including doing nothing. This is the hardest section to get right and the one that makes the biggest difference.

Q1: What are the top 3 products buyers compare you to? (Not "our competitors" — what do BUYERS actually consider before choosing?) Q2: What do buyers do TODAY if they don't buy any tool, including yours? Q3: Why do buyers leave those alternatives and consider you instead? Q4: What can you do that NO alternative can do? (Features, approach, delivery model) Q5: What category do you compete in? (If creating a new category, name it) Q6: Are you in an existing market, new market, or resegmenting an existing market?

Store as: alternatives_primary[], alternatives_status_quo, alternatives_why_leaving, unique_capabilities, category, market_type

Confirm Section 3: Show answers. "Look good? (yes / edit / skip for now)"

---

Step 5: Voice & Tone (5 Questions)

Why this section: Consistency across channels. Every writing skill reads this before producing copy. Without it, outputs sound like AI. With it, they sound like you.

Q1: Pick 3–5 words that describe your brand voice. Q2: Does your tone shift when talking to different personas? Describe the shifts. Q3: Any language preferences? ("We say X, not Y") Q4: Any words or phrases you NEVER use? (Jargon, clichés, competitor language) Q5: Paste a paragraph of copy that perfectly captures your voice.

Store as: voice_attributes[], tone_shifts, language_preferences, forbidden_phrases, tone_example

Confirm Section 4: Show answers. "Look good? (yes / edit / skip for now)"

---

Step 6: Market Context (4 Questions)

Why this section: The "why now" that makes launches feel inevitable rather than arbitrary. Without this, GTM briefs lack urgency and messaging lacks a market narrative.

Q1: How mature is your market? (Nascent / Emerging / Growing / Mature) Q2: What macro trends or forces make your solution relevant NOW? Q3: Complete: "Buyers need [your product] NOW because _____" Q4: What's the bigger story about where the market is going? (2–3 sentences)

Store as: market_maturity, macro_forces, why_now, market_narrative

Confirm Section 5: Show answers. "Look good? (yes / edit / skip for now)"

---

Step 7: Proof Points Registry (3 Questions)

Why this section: Claims without evidence get cut by legal and ignored by buyers. This is the approved claims registry every battlecard and launch brief pulls from.

Q1: What metrics can your team confidently cite? (With sources if possible) Q2: Any customer quotes, testimonials, or case study results you can reference? Q3: Any claims your team is NOT allowed to make? (Unverified, exaggerated, legally risky)

Store as: approved_metrics[], proof_points_quotes, forbidden_claims

Confirm Section 6: Show answers. "Look good? (yes / edit / skip for now)"

---

Step 8: Generate and Save Brain File

Once all 6 sections are confirmed:

• Populate all 31 variables from wizard answers
• Write to /foundation/brain.md
• Delete /foundation/.brain-draft.md if it exists
• Confirm to user:

✅ Your brain is live at /foundation/brain.md

Every PMM skill in this ecosystem will now read from it automatically.
ICP, positioning, alternatives, voice, proof points — captured once, used everywhere.

Want to see your brain file? (yes / no)

---

Step 9: Post-Setup Routing (Role × Focus)

After brain is confirmed live, ask ONE question:

> "What's your primary focus right now? I'll point you to the right skills." > > 1. Launching a product or feature > 2. Fixing or refreshing our positioning > 3. Building competitive intelligence > 4. Understanding our buyers better > 5. Enabling the sales team > 6. Planning the quarter (OKRs, priorities)

Then surface 2–3 skills matched to their answer:

Path 1 — Launch:

• go-to-market-strategy — tier the launch, build the strategy brief
• hs-pre-mortem — stress-test before committing resources
• hs-gaccs-brief — campaign brief with channels and messaging

Path 2 — Positioning:

• hs-positioning-messaging — full positioning BUILD or AUDIT
• hs-alternatives-map — map what buyers actually compare you to
• hs-value-prop-statements — persona-specific value props

Path 3 — Competitive:

• hs-competitive-battlecard — build or refresh a battlecard
• hs-ci-stakeholder-briefing — exec-level competitive newsletter
• hs-alternatives-map — named alternatives framework

Path 4 — Buyers:

• hs-buyer-personas — buying committee map and messaging personas
• hs-icp — build or pressure-test your ICP
• hs-interview-summary — synthesize customer discovery

Path 5 — Sales enablement:

• hs-value-prop-statements — persona-specific statements for decks and talk tracks
• hs-competitive-battlecard — objection handling and attack angles
• hs-stakeholder-maps — map champions, blockers, and internal power

Path 6 — Quarterly planning:

• hs-brainstorm-okrs — build or pressure-test PMM OKRs
• workflow-orchestrator — chain multiple skills into a full program
• hs-prioritization-frameworks — score and tier your initiative backlog

For senior operators and VP-level users, also surface:

• workflow-orchestrator — for any multi-skill program
• hs-ci-stakeholder-briefing — for exec-level competitive updates

---

Outputs

• Files written: /foundation/brain.md — 6 sections, all 31 variables populated.

All PMM skills read this. /foundation/.brain-draft.md — partial state written if wizard is interrupted, deleted once full brain is written.

• Chat output format: Section-by-section confirmation → final brain summary →

success message → post-setup routing (Step 9). Each section shows answers and asks for confirmation before proceeding.

• External side effects: If legacy v1.x files in .agents/, migration is offered

with field-by-field confirmation. Old files are not deleted — user keeps them.

---

Verification

| Check | Standard | Pass = | |---|---|---| | All 6 sections populated | No section is blank or Placeholder | Yes | | Answers are specific | No vague values — ICP has company size, industry, role | Yes | | Alternatives named | At least 2 named alternatives + status quo defined | Yes | | Proof points quantified | At least 1 metric with a number and source | Yes | | Brain file written | /foundation/brain.md exists and is parseable | Yes | | Post-setup routing ran | User received skill recommendations matched to focus | Yes |

---

Do Not Use For

• One-time positioning work → hs-positioning-messaging (reads brain + generates output)
• Competitive analysis → hs-competitive-battlecard
• Buyer research → hs-buyer-personas
• OKR planning → hs-brainstorm-okrs
• Multi-skill programs → workflow-orchestrator (runs skills in sequence using brain)
• Direct file edits without Q&A — always edit brain through this skill's wizard.

Direct edits bypass validation and can introduce inconsistencies.

---

Operating Rules

• Show value before asking for setup. First-run users see the first-run hook (Step

Pre-flight) before wizard questions. They know what they're building and why before they start. This is the Ramp Glass principle: earn the setup, don't demand it.

• Never run wizard without brain state check. Always check /foundation/brain.md

first. If it exists, offer View/Edit/Rebuild — never auto-launch the wizard.

• Validate specificity at every answer. Reject vague answers inline: "We sell to

businesses" → push back: "Be more specific — what company size, industry, or role?" The brain is only as sharp as the answers that build it.

• Save partial state on any interrupt. If user quits mid-wizard, write

/foundation/.brain-draft.md with all completed sections and current question number. On next run, detect draft and offer to resume from exactly where they left.

• Parse legacy files with confirmation, never silently. Show each extracted field

before writing. User must confirm. Silent migration introduces errors that corrupt the brain without anyone noticing.

• Post-setup routing is mandatory. After every completed brain setup or edit,

run Step 9. A brain without a next action is a brain that doesn't get used.

• Role-surface the right skills. VP and senior operators get workflow-orchestrator

and hs-ci-stakeholder-briefing surfaced alongside path skills. IC operators get individual execution skills. Ask about focus, not title — focus reveals role.

• Brain writes require section confirmation. Each section must be confirmed before

writing. Never write the brain file from unconfirmed answers.

• Never recommend direct file edits. If user wants to edit brain manually, redirect

to --edit-section [1-6] wizard flow. Manual edits bypass validation.

---

Quality Gate

Runs after brain file is written, before Step 9 routing.

| Check | Standard | Pass = | |---|---|---| | First-run hook ran (new users) | Value shown before wizard started | Yes | | All 6 sections confirmed | User confirmed each section before proceeding | Yes | | No placeholder values written | Brain file has no blank or "TBD" fields | Yes | | Specificity checks passed | ICP has numbers, alternatives are named products | Yes | | Brain file written to correct path | /foundation/brain.md exists | Yes | | Post-setup routing completed | Step 9 ran and skills surfaced for user's focus | Yes |

---

Self-Improvement Loop

Before every session:

1. Check if /foundation/brain.md exists — load it silently.
2. Check knowledge/brain/rules.md if it exists — apply confirmed quality patterns.
3. Note current brain section scores if last audit was run — flag degraded sections.

After every completed brain setup or edit:

1. Note which sections users most commonly skip or ask to revisit.

Log to knowledge/brain/hypotheses.md.

1. Note which answers most commonly fail the specificity check.

If same field fails 3+ times across sessions → propose a more specific example prompt for that question. Log proposed chan

…

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: stefanoskarakasis
• Source: stefanoskarakasis/Product-Marketing-Skills
• License: MIT
• Homepage: https://stefanoskarakasis.substack.com/

Install and usage instructions live in the source repository linked above.