Create and assure one LinkedIn contribution

UNTESTED · claim_state: UNTESTED · outcome_maturity: UNTESTED

Turn one valuable idea into a LinkedIn article that survives the stranger test without fabricating proof or promising outcomes you do not own. This capability recipe brings together the complete job: stages, Form contract, evidence gates, human authority, agent boundary, falsifier, and next move.

Recipe book model

Capability = JTBD recipe. This page owns the complete job narrative: everything you need to produce one LinkedIn contribution reliably. The FORMAT (artifact Form with slots) lives at LinkedIn Article Form. An instance is one filled Form produced by running this recipe.

GATE

• Authority: human_authority: Accountable publication owner
• Beneficiary: reader seeking a useful idea and the accountable publisher
• Permitted inputs and actions: supplied idea, evidence, constraints, and surface specifications; one draft article through EXTERNAL_READY
• Prohibited: publication without human approval, fabricated proof, unsupported outcome claims, invented testimonials or metrics
• Stop: stop_rule: Stop for fabricated proof, unsupported claims, surface mismatch, self-approval, or unclear human publication gate.

CHECK

Stranger test — can a cold reader state the central claim in one sentence?
Evidence discipline — are all claims supported or marked as assumptions?
Encounter ≠ outcome — is the distinction visible (e.g., "might try" not "will achieve")?
Falsifier — is the condition that would invalidate this idea stated?

NEXT MOVE

The accountable publication owner reviews the EXTERNAL_READY draft, corrects or approves it, makes the human publication decision, then observes reader response and returns evidence to improve the method.

Before

Prepare idea and evidence

Confirm the central idea is worth sharing, gather supporting evidence, name what remains unknown, and set the publication owner.

Human decision

Publish or revise

AI returns an EXTERNAL_READY draft. The owner reviews, corrects, and makes the final publish decision.

Recommended setup

Ingredient	Best-fit choice	Why
Model	GPT-5.6 Terra · medium reasoning	Strong enough for evidence discipline and cold-reader simulation while balancing cost for iterative drafting.
Tool	ChatGPT or an equivalent private AI chat	The first trial needs supplied evidence and constraints—not live web access, automated posting, or self-approval.
Skill	Marketing (biz-mkt-marketing)	Supplies positioning discipline, evidence boundaries, and the encounter ≠ outcome distinction the model alone does not own.
Target artifact	LinkedIn Article (see Form)	Structured contribution with useful idea, evidence discipline, falsifier, and human publication gate.

Put this to work

Create an EXTERNAL_READY LinkedIn article draft

For the accountable publication owner

Copy this prompt. Paste into Claude, ChatGPT, or any AI assistant. The page context is already loaded — send it and get analysis tailored to your role.

You are assisting an accountable publication owner to create one LinkedIn article. You may draft and refine only from the evidence and constraints I supply. You do not publish, fabricate proof, or approve the work yourself.

CONTRIBUTION CONTEXT

- Central idea or worthwhile change: [state it]
- Evidence available: [what you can cite]
- Evidence boundary: [where learning stops; what remains unknown]
- Target reader: [who should benefit]
- Surface and constraints: [LinkedIn article; professional tone; 1200-1800 words]
- Publication owner: [name or role]

PRODUCTION STAGES

Follow these stages in order:

**PLAN**
- State one useful idea or worthwhile change for the reader
- Distinguish encounter (what reader might experience) from outcome (beneficiary change you own)
- Name the evidence boundary: where does the learning stop? What remains unknown or untested?
- Identify falsifier: what would invalidate this idea or force a rewrite?

**PRODUCE**
- Write the article with a cold-reader test: can a stranger state the central claim in one sentence?
- Use supplied evidence only; mark assumptions as assumptions
- Keep encounter ≠ outcome visible (e.g., "readers might try X" not "readers will achieve Y")
- No fabricated testimonials, metrics without provenance, or invented validation

**ASSURE**
- Stranger test: would a cold reader understand the central claim, evidence boundary, and what's unknown?
- Surface match: does this fit LinkedIn's professional article format and norms?
- Evidence discipline: are all claims supported or marked as assumptions/unknowns?
- Falsifier check: is the condition that would invalidate this idea stated?

**EXTERNAL_READY**
Return the draft article, evidence used, unknowns acknowledged, falsifier statement, and this question: "Is this ready for you to publish, or what needs revision?"

FALSIFIER RULES

Generic language, unsupported engagement promises, surface mismatch, or self-approval means STOP. The publication owner makes the final publish decision and names the human publication gate.

Target artifact

The LinkedIn Article Form at /playbook/artifacts/ defines the required slots. The recipe produces one instance:

• One useful idea — the worthwhile change for the reader
• Encounter ≠ outcome — the distinction is visible (what reader might try vs. what you guarantee)
• Cold-reader test — a stranger can state the central claim in one sentence
• Evidence boundary — where learning stops; what remains unknown
• Human publication owner — named accountable decision maker
• Falsifier — what would invalidate this idea or force a rewrite

Workflow family: Marketing & Positioning. This capability recipe owns the complete job: use it when a valuable idea deserves wider reach and LinkedIn is the chosen channel. The recipe brings together stages, Form contract, evidence gates, human authority, agent boundary, falsifier, and next move. It does not create the idea, manufacture proof, or replace the human publication decision.

Outcome

The reader receives one useful idea with clear evidence boundaries. The publication owner receives an EXTERNAL_READY draft that survives the stranger test and preserves honesty about what is known, unknown, and untested.

Use When

Trigger this job when a valuable idea is ready to share and LinkedIn is the appropriate channel. The accountable publication owner keeps the final publish decision, evidence discipline, outcome claims, and stop authority. An assistant may draft and refine through EXTERNAL_READY only.

Inputs

Bring the central idea, available evidence, evidence boundary (where learning stops), target reader, surface constraints (LinkedIn article format), and the named publication owner. Record unknowns as unknowns. An encounter (reader might try) is not an outcome (beneficiary change you own).

Recipe

Follow the four stages in order:

1. PLAN

• State one useful idea or worthwhile change for the reader
• Distinguish encounter (what reader experiences) from outcome (change you guarantee)
• Name evidence boundary: where does learning stop? What's unknown or untested?
• Identify falsifier: what would invalidate this or force rewrite?

2. PRODUCE

• Write article with stranger test: can cold reader state central claim in one sentence?
• Use supplied evidence only; mark assumptions clearly
• Keep encounter ≠ outcome visible throughout
• No fabricated testimonials, metrics without provenance, or invented validation

3. ASSURE

Run these checks:

• Stranger test — cold reader understands central claim, evidence boundary, unknowns
• Surface match — fits LinkedIn professional article format and norms
• Evidence discipline — all claims supported or marked as assumptions/unknowns
• Falsifier check — invalidation condition is stated

4. EXTERNAL_READY

Return draft article, evidence used, unknowns acknowledged, falsifier statement, and ask: "Is this ready for you to publish, or what needs revision?"

Falsifier rules: Generic language, unsupported engagement promises, surface mismatch, or self-approval means STOP immediately.

Human and AI Control

The owner may permit drafting and refinement through EXTERNAL_READY only. Stop immediately for fabricated proof, unsupported outcome claims, self-approval, surface mismatch, or unclear publication gate. The human publication owner makes the final publish decision.

Output and Handoff

Return the EXTERNAL_READY draft, evidence cited, unknowns acknowledged, falsifier statement, human decision (publish/revise/stop), and publication date if approved. The published article is the instance; this recipe is the how.

Evidence and Stop

Retain the input idea, evidence supplied, draft iterations, ASSURE checklist results, human corrections, publication decision, and any stop events. Stop using the assistant when it fabricates proof, claims unsupported outcomes, approves its own work, or publishes without human gate. Preserve stopped results and recover through manual review.

Reader engagement metrics (views, likes, comments) are interaction signals, not proof that the idea improved a beneficiary outcome. A viral post does not prove value transferred. A quiet post does not prove the idea failed.

Next

If the stranger test fails, preserve the draft, identify which slot missed (central claim unclear, evidence boundary missing, unknowns hidden, falsifier absent), revise that slot, and retest with a different cold reader. If surface mismatch occurs, check LinkedIn article norms and revise. If evidence discipline fails, name which claims lack support and either add evidence or mark as assumption/unknown.

When the contribution is published, observe reader questions and responses. When readers ask about the evidence boundary or unknowns, that's success—it means they trust the honesty. When readers claim outcomes you didn't promise, correct the encounter ≠ outcome framing in the next contribution.

No clinic or booking CTA. This capability recipe is the complete JTBD: job, stages, Form contract, evidence gates, human authority, agent boundary, falsifier, next move. The FORMAT (Form with slots) lives at /playbook/artifacts/. Link recipe ↔ Form both ways; this recipe page owns the whole job narrative. Different recipes may use different Forms.