IB Deck Engine Skill

When to use this skill

Use this skill any time the user wants to build investment banking presentations, including:

• Full pitch decks / discussion materials / board presentations
• Single slides (financial summary, football field, comps table, S&U)
• Sell-side or buy-side mandates
• Fairness opinions
• IC memos with embedded charts
• Pitch books with consistent formatting

Trigger phrases include: "build a pitch deck for [COMPANY]", "create an IB football field", "DCF sensitivity table", "trading comps for [SECTOR]", "sources and uses slide", "board discussion materials", "fairness opinion deck", "[COMPANY] pitch book".

Architecture: JSON Spec → Renderer

This skill uses a separation-of-concerns architecture. The LLM (you) handles content reasoning. A deterministic Python renderer handles pixel placement.

Your job:

1. Pick the right template for the slide type
2. Fill out the JSON spec for that template
3. Call IBRenderer.render_*() methods
4. Save the PPTX

You never write:

• Inches(), Pt(), EMU() values
• Pixel coordinates
• python-pptx shape positioning code
• Manual color hex placement

The renderer guarantees:

• Right-alignment on all numeric columns
• Proportional bar heights based on data ratios
• Tables are real PowerPoint table objects (not text with tabs)
• Reference line labels never overlap titles
• Source text always positioned below content, above footer
• Consistent chrome (logo, confidential mark, badge, slide number)

Available templates

Template	Method	Use for
Cover	render_cover	Project cover slide with confidential mark and bank badge
Section Divider	render_section_divider	Full navy background with white section title
Table of Contents	render_toc	Numbered agenda with light blue bands
Executive Summary	render_exec_summary	Blue callout bar + ■ / — bullet hierarchy
Investment Highlights	render_investment_highlights	4-card 2×2 grid with numbered headers
Financial Summary	render_financial_summary	Historical P&L with bold subtotals + italic % rows
Stacked Bar + Table	render_stacked_bar_table	Moelis revenue-by-segment pattern
Dual Chart	render_dual_chart	GS-style two charts side by side with CAGR labels
Football Field	render_football_field	Valuation summary with category pills + range bars
Sensitivity Grid	render_sensitivity	WACC × Terminal Growth → share price grid
Sources & Uses	render_sources_uses	LBO capital structure with balance check
Trading Comps	render_trading_comps	Peer multiples table with target highlighted

See reference/template-catalog.md for full schema details and reference/examples/ for working JSON specs.

Workflow

Step 1: Confirm what the user wants

If the user says "build an IB deck for ADUS", confirm:

• Single company sell-side pitch? Buy-side analysis? Fairness opinion?
• How many slides? (Full pitch book — ~14 slides using the 12 template types with section dividers reused — or specific slides?)
• Any specific data sources to use?

Step 2: Get the data into the master JSON schema

The plugin does not automate SEC EDGAR extraction in this version. You have three options for getting data into the master JSON schema:

1. Ask the user to paste financials or supply a CSV / existing JSON
2. Write a short extraction script using edgartools as a one-off (install separately: pip install edgartools[ai])
3. Use the companion case study at github.com/gorajing/ib-deck-engine which has a worked ADUS extraction as a reference

The target JSON schema is documented in reference/template-catalog.md. Whatever path you take, the output of this step should be a single JSON file that every subsequent template call reads from — the "single source of truth" pattern prevents cross-model desync.

Step 3: Pick templates and fill specs

For each slide, pick the matching template and write a JSON spec:

from ib_deck_engine import IBRenderer

r = IBRenderer()

r.render_financial_summary({
    "title": "Consistent Revenue Growth with Expanding Profitability",
    "section_header": "Historical Financial Summary ($ in thousands)",
    "headers": ["Metric", "FY2023A", "FY2024A", "FY2025A", "3Y CAGR"],
    "rows": [
        {"label": "Revenue", "values": ["1,058,651", "1,154,599", "1,422,530", "14.4%"], "style": "bold"},
        {"label": "% Growth", "values": ["11.3%", "9.1%", "23.2%", ""], "style": "pct"},
        {"label": "EBITDA", "values": ["105,082", "116,221", "155,027", "21.4%"], "style": "highlight"},
        {"label": "% Margin", "values": ["9.9%", "10.1%", "10.9%", ""], "style": "pct"},
    ],
    "source": "Source: SEC EDGAR. FY2025 10-K.",
})

Step 4: Save and verify

r.save("output.pptx")

Then verify by reading the output: confirm slide count, check that source text isn't cut off, verify no text wrapping issues.

Critical rules

Action titles, not generic headers

WRONG: "title": "Financial Summary" RIGHT: "title": "Consistent Revenue Growth with Expanding EBITDA Margin Expansion"

Every slide title must be a declarative thesis, not a label. This is the #1 IB formatting standard. The audience should understand the slide's point from the title alone.

Right-alignment is automatic — don't override

The templates already right-align numeric columns. You don't need to specify alignment in the JSON spec. If you find yourself wanting to override, you're doing it wrong.

Use real numbers, sourced

Every number on every slide must trace back to a source file (the master JSON, the DCF model, the LBO model). Never invent numbers. If you don't have data, ask the user.

Source citations on every slide

Every slide spec should include a source field. Format: "Source: SEC EDGAR via edgartools. FY2025 10-K filed February 24, 2026."

Cross-model consistency

When building both DCF and LBO models for the same company, document any assumption differences explicitly:

• Tax rate: DCF uses effective rate, LBO uses 21% statutory (write WHY)
• Revenue growth: LBO can be more conservative (write WHY)
• These differences should be NOTED, not hidden

Common patterns

Pattern 1: Full pitch book (~14 slides using 12 template types + reused section dividers)

r.render_cover({...})
r.render_toc({...})
r.render_exec_summary({...})
r.render_investment_highlights({...})
r.render_section_divider({"title": "Financial Performance"})
r.render_financial_summary({...})
r.render_stacked_bar_table({...})  # revenue by segment
r.render_dual_chart({...})  # revenue + EBITDA
r.render_section_divider({"title": "Valuation"})
r.render_football_field({...})
r.render_sensitivity({...})
r.render_trading_comps({...})
r.render_section_divider({"title": "LBO Analysis"})
r.render_sources_uses({...})
r.save("project_alpine.pptx")

See reference/examples/full_deck.py for a complete worked example that uses every template and reuses render_section_divider for the three part breaks.

Pattern 2: Single slide quick build

For users who want one slide, ask which template fits and just call that one method.

Pattern 3: Update existing deck

The renderer creates new slides; it doesn't modify existing PPTX. For "update slide X with new data," generate a fresh slide and ask the user to copy it into their deck.

Troubleshooting

"Right-alignment isn't working" — You shouldn't need to set this. The renderer does it automatically for all numeric columns in financial_summary, trading_comps, sensitivity, and sources_uses templates.

"Bars aren't proportional" — Check the values array in your stacked_bar_table or dual_chart spec. The renderer computes heights as value / max_value * chart_height. If bars look wrong, your input data is wrong.

"Source text is cut off" — The renderer dynamically positions source text below content. If it's cut off, your content is too tall. Reduce the number of rows in the table or use a shorter chart.

"Reference line labels overlapping" — The football field template has built-in collision detection. If two reference labels would overlap horizontally, the second one bumps to a second row automatically.

"openpyxl recovery dialog (when generating Excel separately)" — If you're writing your own Excel file alongside the slides, use xlsxwriter for new file creation. openpyxl is only for reading/editing existing files. The .pptx rendering in this plugin uses python-pptx, not openpyxl, and is not affected by this issue.

Examples

See reference/examples/ for working JSON specs you can copy as starting points:

• financial_summary.json — historical P&L table
• football_field.json — valuation summary with reference lines
• stacked_bar_table.json — stacked bars + data table (revenue by segment)
• sources_uses.json — LBO capital structure
• trading_comps.json — peer multiples table
• dual_chart.json — side-by-side bar charts with CAGR labels
• sensitivity.json — WACC × TGR grid
• full_deck.py — complete worked deck using every template

Reference docs

• reference/template-catalog.md — All 12 templates with full JSON schemas
• reference/style-guide.md — Color palette, typography, formatting conventions
• reference/architecture.md — Why the renderer-first architecture produces more repeatable output than prompt-guided spatial code