Mitchell Agentic Sprint

AI-led 6-step sprint that walks an AI builder through validating one idea — vertical, buyers, framework, positioning — to a sales deck, outreach plan, and pre-seed investor deck. Adversarial by default. Output is artifacts, not product.

View project on GitHub

SPEC-009-A: Artifact Template Format

Field Value
Specification ID SPEC-009-A
Parent ADR ADR-009
Version 1.0
Status Active
Last Updated 2026-05-09

Overview

Defines the structure of the three Step 6 artefacts (sales deck, customer outreach plan, pre-seed investor deck) and the rules commands/sprint-step-6.md follows when populating them.

The templates live at:

  • docs/templates/sales-deck.md — 6–8 slides for customer conversations
  • docs/templates/outreach-plan.md — re-engagement messages + new-contact target list + outreach sequence
  • docs/templates/investor-deck.md — 12 slides for pre-seed conversations

/sprint-step-6 reads each template, populates it with the user’s evidence from Steps 1–5, and writes the populated artefact to .sprint/artifacts/. The original templates in docs/templates/ are never modified — they are read-only scaffolds.

Universal rules across all three artefacts

  1. Every claim cites evidence. Each slide / section that makes a claim about buyer pain, market dynamics, traction, or moat must reference the source — interview quotes (with pseudonym + interview number), expert framework citations (with source document), competitor verbatim quotes (with competitor name).
  2. No padding when evidence is missing. If a slide’s required evidence is missing, the slide is marked [EVIDENCE MISSING: {what's needed}] rather than fabricated. The user resolves the gap before the artefact is presentation-ready.
  3. Verbatim quotes only. When citing buyers, use the verbatim language from the interview transcripts — do not paraphrase. Concept §6.
  4. Slide-per-section structure. Each slide is a level-2 markdown heading with the slide number and title. Slide content sits below; speaker notes go in a fenced > Speaker notes: block at the end of each slide.
  5. Anti-sycophancy preamble in the artefact frontmatter. Each populated artefact opens with a one-paragraph note for the user: “These artefacts are populated from Sprint state. Where evidence is marked missing, fill it in before presenting. Do not pad.”

Sales deck template

Path: docs/templates/sales-deck.md Slides: 6–8 Audience: prospective customers in the user’s chosen ICP

Required slide order

# Slide Required evidence
1 Buyer pain (verbatim quote opens) One verbatim quote from the strongest theme in Theme Map, attributed to a pseudonymised buyer
2 Problem stated precisely Plain-language problem statement; cite the recurrence (number of interviews) and costly-action signals
3 Differentiated solution One paragraph + the positioning whitespace from Step 5; cite the verbatim language no competitor uses
4 Validation evidence Theme Map summary + 2–3 strongest costly-action signals + count of interviews logged
5 Framework basis (optional, skip if it doesn’t add credibility) Expert framework citation showing the principle being adopted
6 Pricing or offer Pricing landed in Step 4/5 work; cite the unit-economics reasoning if applicable
7 Next step Specific named ask: book a 30-min walkthrough, take a paid pilot, intro to a peer
8 Appendix (optional) Costly-Action Signal list with anonymised buyer detail

The deck doubles as a validation-proof document: the buyer interview evidence sits alongside the pitch.

Outreach plan template

Path: docs/templates/outreach-plan.md Sections: 3

Required sections

  1. Re-engagement messages for the buyers interviewed in Step 3. Auto-generated from the Interview Log:
    • For each interviewee with a costly-action signal: a personalised re-engagement message in the spirit of “here’s what I built based on what you told me”, quoting the buyer’s own verbatim phrase from the interview
    • One message per costly-action signal, with the buyer’s pseudonym and the verbatim quote inline
    • The user replaces pseudonyms with real identifiers before sending
  2. New-contact target list. Drawn from the user’s Vertical Stack ICP definition + Step 4’s expert framework + Step 5’s positioning whitespace:
    • 15–25 new buyer profiles matching the ICP
    • For each: where to find them (community / role / company-size pattern), why they fit, the suggested first-touch hook (anchored to the validated theme)
  3. Outreach sequence. A 3-touch cadence drafted around the validated problem (not a vague pitch):
    • Touch 1 — first contact, behaviour-anchored
    • Touch 2 — follow-up, references the validated theme
    • Touch 3 — closing ask, drawn from the sales deck’s “next step”

Investor deck template

Path: docs/templates/investor-deck.md Slides: 12 + appendix Audience: pre-seed investors

Required slide order

# Slide Required evidence
1 Problem Buyer pain in verbatim words; recurrence count
2 Why now Specific 2026-AI-era reasoning, not generic “AI is hot”
3 Solution Plain language; what the user built / will build
4 Product Working demo screenshot reference (if missing, marked) + 1–2 user-experience highlights
5 Market Bottom-up sized from ICP × buyers-reachable, not top-down TAM
6 Business model Priced against actual buyers; cite the buyer who paid for the workaround
7 Traction Named buyers + costly-action signals + interview count + LOIs / pilots
8 Competition Competitor cluster map from Step 5 + the cluster the user does not occupy
9 Moat Specific lock-in mechanism: workflow / data flywheel / regulated-vertical access. Not “proprietary AI”.
10 Team Builder Profile from Step 1 — unfair advantages + relevant experience
11 Ask Specific use of funds, runway extension, milestones
12 Appendix Inference economics (cost per query × queries per buyer × buyer count); evals plan; buyer-interview tape pointer

2026-AI-era non-negotiables

The Step 6 command bakes these in. Each is a checklist item the populated deck must satisfy or explicitly mark as missing:

  • Working demo (slide 4)
  • Real moat — workflow lock-in / data flywheel / regulated-vertical access (slide 9; “proprietary AI” is rejected)
  • Evals — what success looks like, how it’s measured (appendix)
  • Inference economics — gross margin under inference cost (appendix)
  • Buyer-interview tape pointer — .sprint/interviews/ directory + Theme Map (appendix)

When a check is missing, the slide is marked [EVIDENCE MISSING: {what's needed}]. The user fills the gap before sending the deck.

Where artefacts are written

/sprint-step-6 writes populated artefacts to:

  • .sprint/artifacts/sales-deck.md
  • .sprint/artifacts/outreach-plan.md
  • .sprint/artifacts/investor-deck.md
  • .sprint/artifacts/audio-overview.mp3 (only if NotebookLM bolt-on is configured; concept §4.6 calls this out as a supporting artefact)

Their paths are recorded in .sprint/sprint.md Artifacts section so subsequent Sprints / continue-quest commands can locate them.

Acceptance criteria

This spec is fulfilled when:

  • Three template files exist at docs/templates/{sales-deck,outreach-plan,investor-deck}.md with the slide / section structure above
  • Each template uses placeholder `` markers that the Step 6 command fills with state from .sprint/sprint.md
  • The Step 6 command writes populated artefacts to .sprint/artifacts/ without modifying the templates
  • Slides where evidence is missing are marked [EVIDENCE MISSING: ...] rather than fabricated
  • Each artefact opens with the anti-sycophancy preamble note for the user
  • The investor deck satisfies (or explicitly marks missing) all five 2026-AI-era non-negotiables