draft-pitch

Draft Pitch — One-pager / executive summary / deck-equivalent

You are drafting a longer-form pitch document. Three shape options; choose based on the user's stated need:

• one-pager (~500-700 words) — single-page document for sharing with potential funders, partners, or stakeholders
• exec-summary (~300-500 words) — denser front-matter for a longer proposal or memo
• deck-content (~600-1000 words, structured into slide-shaped sections) — text content for a pitch deck (slide titles + 2-4 bullet points per slide; user assembles the actual visual deck)

All three are LONGER than elevator-pitch (~150 words) but SHORTER than a full proposal. They serve as the first-touch document a reviewer sees.

Foundational rules

1. Grounded. Every substantive claim must trace to repo source. Quantitative claims (user counts, revenue, traction metrics) must be backed by repo evidence (a stats file, a metrics dashboard reference, a README line, etc.) or labeled as estimates/projections.
2. Not overpromising. No superlatives. No "the only" without backing. No vague-quantity claims without enumeration. The validators will catch these; pre-edit so they don't have to.

Audience modes

investor (default for one-pager / exec-summary)

For a VC, angel, or accelerator evaluating commercial potential.

Required sections:

1. The opportunity (problem + market context, 1 paragraph)
2. What we're building (solution shape, 1-2 paragraphs)
3. Why now (timing — what's changed that makes this viable, 1 paragraph)
4. Who it's for (target customer in measurable terms, 1 paragraph)
5. Differentiation (vs named alternatives — concrete, not superlative, 1 paragraph)
6. Stage and traction (current metrics, growth trajectory, what's been validated, 1 paragraph)
7. The team (only if user-provided; do NOT invent)
8. Business model (how money flows, 1 paragraph)
9. What we're asking for (only if user-specifies; do NOT invent)

grant

For an academic, foundation, or government grant application.

Required sections (most grants want):

1. Project summary (1 paragraph)
2. Problem and significance (the public/scientific/social good case)
3. What we'll do (specific activities)
4. What we'll produce (deliverables)
5. Why we can do it (capability — only what traces to source)
6. Budget and timeline (only if user-provided)
7. How we'll measure success

partner

For a potential integration / co-marketing / strategic partnership.

Required sections:

1. What we offer (capabilities)
2. What we're looking for (the ask)
3. Mutual fit (why this partnership specifically)
4. Integration shape (technical: APIs, data flows, joint go-to-market shape)
5. Stage and timing

Shape: deck-content

When --shape deck-content, structure the output as slide-shaped sections.

Each "slide" is:

• A 4-8 word slide title (H2)
• 2-4 bullet points (each ~10-15 words)
• An optional "speaker note" paragraph below

Standard deck shape (8-12 slides):

1. Title slide (project name + one-line description)
2. The problem (what's broken in the world)
3. The solution (what we built)
4. How it works (architecture / flow / mechanism)
5. Market / audience (who needs this)
6. Differentiation (vs named alternatives)
7. Traction (metrics, milestones, validated outcomes)
8. Business model (if applicable)
9. Team (if user-provided)
10. Ask + timeline
11. Closing slide

Skip slides without source backing; don't invent.

Workflow

Step 1: Determine shape and audience

If --shape and --audience were both passed, proceed. Otherwise ask:

What shape?

  1. One-pager (~500-700 words) — single-page share
  2. Executive summary (~300-500 words) — proposal front-matter
  3. Deck content (~600-1000 words) — pitch deck text

Who's it for?

  1. Investor (VC / angel / accelerator)
  2. Grant (academic / foundation / government)
  3. Partner (integration / co-marketing / strategic)

Step 2: Confirm output target

Offer:

• PITCH.md / EXEC-SUMMARY.md / DECK.md at repo root
• docs/pitch/<audience>-<shape>.md (in-repo persistence)
• Custom path

Step 3: Read source

In parallel:

• README.md
• package.json / pyproject.toml / similar manifest
• docs/vision.md (if present) — primary source for problem/solution/audience
• docs/architecture.md (if present) — for solution shape
• docs/roadmap.md (if present) — for stage and trajectory
• docs/status.md (if present) — for current metrics
• docs/decisions/*.md (if present) — for differentiation, why-this-choice context
• Recent commits via git log --oneline -20 — for stage signal
• For investor pitch: optionally search web for the named alternatives the user mentions (verify they exist; don't invent competitors)

Step 4: Surface gaps before drafting

Before writing, surface to the user:

Source coverage for {shape} / {audience}:

  ✓ Have: {list of sources found}
  ✗ Missing: {list of sources not found, with what they would have provided}

Substantive gaps that would weaken this pitch:
  {list — e.g., "no team section because no team info in repo", "no traction
   metrics because no docs/status.md or analytics file", etc.}

How to proceed:
  1. Draft what's possible from the available sources; flag gaps as TODO inline
  2. Pause; you fill in the gaps first, then come back
  3. Draft with placeholders; you fill in before sending

Step 5: Draft section by section

For each section:

1. Pull source content
2. Convert to audience-appropriate prose
3. Cross-check every claim against source
4. If claim has no source: drop, soften, or mark as TODO/placeholder (never invent)

Step 6: Run validators

python3 ${CLAUDE_PLUGIN_ROOT}/scripts/check-grounding.py {output-path} --repo <repo-root> --json
python3 ${CLAUDE_PLUGIN_ROOT}/scripts/check-overpromise.py {output-path} --json

For investor pitches especially: any superlative or vague-quantity claim that the validators flag is one a real investor will ask about in due diligence. Better to fix it now.

Step 7: Surface what you did

Pitch written to {path}.

Shape: {one-pager | exec-summary | deck-content}
Audience: {investor | grant | partner}
Word count: {N}
Sources used: {list}
Sections with source backing: {N} / {M}
Sections with TODO/placeholder (you need to fill these): {list}
Validators: grounding {pass|N flagged}, overpromise {pass|N flagged}

Specific things to verify before sending:
  - {list of claims that needed user judgment to keep}
  - {team/financial/legal claims user must provide}

Rendering rules

• Specific verbs and nouns. "rad-explain runs a grounding validator that flags untraceable claims" > "rad-explain provides advanced validation capabilities."
• Concrete numbers when source-backed. Range them honestly when uncertain ("v0.1.0 with 5 skills + 2 validators", "in active use across [N] projects this month").
• Comparison-shaped differentiation. "Unlike X, we do Y because Z" — not "the best at A."
• Cite alternatives by name. "ChatGPT will puff your project; rad-explain flags the puff" is better than "other tools puff your project."
• One claim per sentence in the heavy-traffic sections (Solution, Differentiation, Traction).

Cross-plugin notes

• For a much shorter pitch (~150 words / 30 seconds spoken), use elevator-pitch.
• For a full project narrative, use narrate-project in this plugin.
• For interpreting a specific repo file or plan, use explain-document in this plugin.
• Both validators live in this plugin's scripts/ dir.