Executive Slide Builder

This skill governs how to construct the narrative, select slide structures, and write executive-quality content. Actual .pptx file generation is handled by the metis-pptx skill — once the story is locked, this skill produces a structured build brief that metis-pptx can execute directly.

Scope boundary — read this first: This skill produces narrative content and a build brief. It does NOT produce .pptx files, Python scripts, or any slide-generation code. The .pptx build is owned entirely by the metis-pptx skill. When this skill's workflow is complete, stop and present the build brief to the user. Do not write python-pptx code, do not import pptx, do not create build_deck.py, and do not attempt to generate a PowerPoint file. If the user asks for the actual deck, tell them to invoke metis-pptx with the build brief — or offer to do so on their behalf by calling that skill. The division is: this skill decides what to say; metis-pptx decides how to render it.

What you get from this skill:

1. A content map — the key messages, proof points, and what stays off the slide
2. A slide-by-slide outline — one job per slide, sequenced for narrative flow
3. Executive-quality draft copy — concise, quantitative, skimmable
4. A build brief — a structured handoff to metis-pptx for .pptx generation

The agent should not invent the story from scratch. It should:

• Gather the relevant context from source materials
• Infer the required narrative structure
• Reuse existing slide architectures from the asset library
• Draft concise, executive-style copy
• Iterate by tightening and simplifying rather than rebuilding

Core Operating Principles

1) Context first, then content

Before drafting any slide, identify and review the relevant source materials. The user's ask is rarely self-contained — real guidance lives in:

• Uploaded briefs, outlines, or transcripts
• Existing decks or meeting notes
• Spreadsheets or models (for numbers)
• Brand templates or visual frameworks

If source material is referenced indirectly, look for it rather than defaulting to generic content.

2) Separate content truth from presentation truth

• Content source of truth: documents that determine facts, metrics, sequencing, recommendations, and claims
• Presentation source of truth: templates and framework slides that determine tone, layout, density, and visual structure

A style guide does not override facts. Content sources do not invent new slide architectures when a reusable one already exists.

3) One slide, one job

Every slide has a single job. Before drafting, name it explicitly. Anything that belongs to another section gets cut or moved — not squeezed in.

4) Reuse structure before inventing structure

Always search the shared asset library for an existing layout before creating a new one. The library contains: slide skeletons, roadmap layouts, comparison pages, phased transformation pages, business-case structures, summary pages, KPI cards, and framework visuals.

Shared asset library:

• G drive: G:\Shared drives\Knowledge Management\New Brand Assets\PPT Assets\Metis_Timesaver Slides.pptx
• Google Slides: https://docs.google.com/presentation/d/1KtfpTOi0CTdySmzN4usaiHJJhXF63KoS/edit

Map the user's outline onto the closest fitting structure from this deck before writing any slide copy.

The Five-Layer Model

Content exists at five layers. The goal in any executive-facing output is always to reach Layer 1. This model is the diagnostic for mode detection and the reference for naming, framing, and quality checks throughout the workflow.

C-Suite audiences live at Layers 1 and 2. They do not need Layers 4 and 5. Layer 3 may appear as supporting context but should never lead.

The executive test for every piece of content: "Can the most senior leader in the room not know what [system name] means and still understand why this matters?" If the answer is no, the content is not yet at the right level.

Mode Detection

After reading source materials (during the Context Retrieval Protocol), assess where the content sits on the Five-Layer Model:

• Elevation Mode — activates when source material is dominated by system names, acronyms, product/vendor names, technical architecture language, or when capabilities are described by what they are rather than what they do for the business. If most content sits at Layers 3–5, elevation is needed. State this to the user:

  "The source material is primarily at the systems and product level — I'll run the elevation workflow to translate it to executive language before building the narrative."

• Narrative Mode — activates when the content is already at Layers 1–2 and the user's challenge is sequencing, emphasis, or story structure. State this to the user:

  "This content is already at the business outcome level — I'll move straight to narrative structure and slide sequencing."

If the assessment is wrong, the user corrects and the agent adjusts. No extra decision point, no friction.

Elevation Mode flow: Client Setup (4 Anchors) → E1: Audit → E2: Separate → E3: Value Stream Map → E4: Name → E5: Time Horizons → E6: Headline → then into the Required Workflow (Steps 1–10) → Completeness Check → So What Test → Build Brief → metis-pptx

Narrative Mode flow (unchanged): Context Retrieval → Steps 1–10 → Quality Bar → Build Brief → metis-pptx

Context Retrieval Protocol

At the start of every session, actively locate the relevant context. Do not assume context from a previous session carries over.

Priority 1 — User-provided files in this session

Read any files the user has uploaded or referenced first — briefs, outlines, transcripts, models, decks. These are the most authoritative.

Priority 2 — PowerPoint template for this session

Before building any slide, confirm which template to use. Templates change per client and engagement.

Ask:

"Please share the PowerPoint template for this project — a file upload, Google Slides link, or file path. I'll use it to govern layout, colors, fonts, and slide density for everything I build."

If the user doesn't provide one, fall back to the shared asset library — but flag this rather than defaulting silently.

When a template is provided:

• Read it before drafting
• Extract: color palette, fonts, title treatment, body text sizing, header/footer conventions, box and card structures, recurring layouts
• Treat those conventions as binding — do not introduce layouts, colors, or font styles not present in the template
• Flag explicitly if something the user wants can't be achieved within the template's constraints

Priority 3 — Shared drive (for slide structures and brand assets)

For slide structure and reusable frameworks, consult: G:\Shared drives\Knowledge Management\New Brand Assets\PPT Assets\Metis_Timesaver Slides.pptx

This is the source of truth for layouts and visual structure. Use it before inventing a new slide architecture.

Priority 4 — Ask before guessing

If no relevant context can be found, tell the user specifically what is missing and ask for it. Never silently proceed with assumed or cached context.

How to Open a Session

When the user invokes /executive-slide-builder or signals a narrative or elevation challenge, start here (if not already provided):

1. Template: "Please share the PowerPoint template for this project (file, link, or path). If you don't have one, I'll use the shared asset library."

2. Project context: "What's the client or project name, and who is the audience for this section?"

3. Read source materials and detect mode. After receiving context, read all source documents and assess the Five-Layer Model to determine Narrative Mode or Elevation Mode. State the mode to the user.

Wait for context before building a content map or drafting any slide. The template governs everything visual; the audience governs everything about emphasis and tone.

Elevation Pre-Processing

This section runs only in Elevation Mode. In Narrative Mode, skip directly to the Required Workflow.

The purpose of elevation is to translate source material written at the systems or technical level into executive-ready content that leads with business value, not technology. It is client-agnostic — the four anchors make it specific to each organization.

Client Setup: The Four Anchors

Complete these before running any elevation step. Everything downstream depends on getting these right. Populate them from source documents, stakeholder interviews, or existing strategy materials.

For detailed templates and formats for each anchor, read references/elevation-reference.md, Section 1.

Anchor 1 — The Value Stream Frame How the organization describes the journey from winning a customer to keeping one. Find their version — it is almost always in a strategy deck, annual operating plan, or investor presentation. Document the zones, what each covers, and the primary metric for each. Cross-cutting capabilities (platforms, data infrastructure, AI foundations) span all zones and sit as foundation, not inside any single column.

Anchor 2 — The Portfolio Proof Point The single most powerful number: the total return on the technology portfolio — not any one system. This is the headline that gives every capability claim credibility. Document as: [Total benefit] from [portfolio/program name] at [total investment]. If no proof point exists yet, identify the closest proxy.

Anchor 3 — The Naming Convention The implicit naming style for digital and AI capabilities. Pick one or two patterns that fit the client's voice and apply them consistently. Inconsistent naming across slides is the most visible signal that content is not yet executive-ready. See the reference file for common naming patterns.

Anchor 4 — The Time Horizon Labels The client's language for investment stages. Executives think in horizons, not quarters. Find these from their STRAP, AOP, or board materials. If the client already uses specific horizon language, use their words exactly. See the reference file for common formats.

Step E1 — Audit the Source Material

Read all source documents and extract into a working inventory. Do not write anything yet.

What to collect:

• Every system name and acronym
• Every AI product or tool name
• Every digital program or initiative name
• Every stated business outcome with a number attached
• Every stated business outcome without a number
• Every investment figure tied to a program or initiative

Flag immediately:

• Capabilities that have systems but no stated business outcome
• Capabilities that have outcomes but no named system or program
• Any part of the client's value stream that has nothing mapped to it — these are genuine gaps that become questions or risks in the narrative

The gaps are often as important as the content. An empty column on an executive slide asks a question the audience will ask out loud.

Step E2 — Separate "What It Is" from "What It Does"

For every item in the inventory, write two sentences:

1. Technical sentence: What is this system or product, and how does it work?
2. Business sentence: What problem does it solve, and what does it change for the business?

Only the business sentence will appear in C-Suite materials. The technical sentence is the internal reference for validating accuracy.

If the business sentence cannot be written, the capability is not yet ready for executive consumption. Return to source documents or flag for a stakeholder conversation.

Step E3 — Apply the Value Stream Frame

Map every capability to a position in the client's value stream (Anchor 1).

For each capability ask:

• Where in the value stream does this create value?
• Which primary metric does it move?
• Does the client already have a named column for this?

Cross-cutting capabilities span all zones and sit below as a foundation layer, not inside any single column.

Step E4 — Name at the Right Level

Apply the naming convention (Anchor 3) consistently.

Rules:

1. Drop all acronyms unless universally known at the executive level
2. Apply the client's naming convention consistently
3. Name to the outcome, not the feature — the name should imply what the capability achieves, not what it is
4. Check consistency across all slides — same capability, same name
5. Watch for premature naming — a building capability gets a name; a live capability gets a name and a number

Step E5 — Map to Time Horizons

Using the client's horizon labels (Anchor 4), assign every capability to the correct time window.

Critical discipline: never conflate deployment with value realization. These are different events, often one year apart. Showing both — correctly labeled — demonstrates that the author understands the difference between activity and outcome.

Step E6 — Build the Headline First

Before arranging capability content on a slide, write two things:

1. The headline number (from Anchor 2) — the total portfolio return. A single aggregated figure contextualizes every individual capability below it. Without it, the slide is a list. With it, the slide is a business case.

2. The framing sentence — one sentence that tells the executive what the slide is arguing, not just what it contains.

   • Not: "This slide shows our digital capabilities across the value stream"
   • Yes: "[Client name]'s digital investments are converting operational complexity into measurable commercial advantage — protecting revenue, expanding margin, and differentiating the customer experience."

Customize the framing sentence to the client's strategic priorities.

After completing E6, proceed to the Required Workflow (Step 1) with the elevated content as input.

Required Workflow

Step 1: Name the slide's job

Before drafting anything, answer:

• What is this slide's single job in the deck?
• Is this one slide or a section of slides?
• What does the audience need to believe or decide after seeing it?
• Is there an outline that governs sequence?

If an outline exists, it governs section flow unless stakeholder feedback explicitly overrides it.

Step 2: Gather and read the source material

Actively find and review: brief, business case, outline, transcript or feedback notes, models or spreadsheets for numeric support, prior slides, reusable frameworks from the asset library.

If context is incomplete, look for it in connected files or referenced folders before asking the user.

In Elevation Mode, the translation table from pre-processing serves as the primary content input for this step.

Step 3: Build a content map

Extract from the source material:

• Key messages (what must land)
• Major numbers and proof points
• Phase logic or chronology
• Internal vs. external value distinctions
• What stakeholders explicitly want emphasized
• What does not belong on this slide

Use the content map to determine hierarchy and visual prominence before touching any slide structure.

In Elevation Mode, the content map should already be at Layers 1–2. If any Layer 3–5 content remains, send it back through Steps E2–E4 before proceeding.

Step 4: Select the right slide structure

Match the content map to a slide form from the asset library:

• Headline + proof points
• "Three forces" convergence
• Roadmap with phase columns
• Phased build / scale / expand
• Framework layering
• Business case / investment
• Executive summary with chart placeholder
• KPI summary / metric cards

Never force all content into the same format. Pick the structure that serves the message.

Step 5: Draft in executive prose

Write concisely, quantitatively, in active voice, easy to skim.

Rules:

• Short declarative bullets
• Lead with the action or outcome
• Use numbers when available
• One idea per bullet
• No paragraphs unless the user asked for narrative copy
• No jargon unless it appears in source materials
• Prefer "show, don't tell"

Step 6: Let the visual do the heavy lifting

If a framework or chart carries the argument, reduce surrounding text. Bullets should clarify: what is in focus, why it matters, what changes from one phase or slide to the next.

Do not duplicate what the visual already shows.

Step 7: Ground every number and stub every placeholder

All metrics, projections, and quantitative claims must trace back to source material — never invented. If a number is uncertain, flag it rather than guess.

When a final visual is not ready, create the slide shell with a clearly labeled placeholder and stable surrounding copy. This lets the deck move forward without waiting on the visual.

Step 8: Sequence multi-slide sections intentionally

Before drafting individual slides, establish the narrative sequence:

Frame the issue → Explain the model → Show the roadmap → Show the economics → Summarize for executives

Each slide should build on the one before it. If the sequence doesn't hold up without a slide, that's a sign it shouldn't exist.

Step 9: Produce a build brief for metis-pptx — STOP HERE

This is the final output of the executive-slide-builder skill. Once the narrative structure and content are confirmed, produce a structured build brief and present it to the user. This is where this skill's job ends.

Do not proceed to write Python code, generate .pptx files, or invoke python-pptx. The build brief is the deliverable. If the user wants the actual PowerPoint file, tell them:

"The narrative and build brief are ready. Want me to hand this off to metis-pptx to generate the deck?"

If they say yes, invoke the metis-pptx skill with the build brief. Do not attempt to replicate what metis-pptx does — it owns brand colors, fonts, layout grids, component code, template files, and the Python build script.

Build brief format:

DECK TYPE: [proposal / in-project deliverable / other]
TEMPLATE: [file path or "shared asset library fallback"]

SLIDE 1
  Title: [slide title]
  Job: [one sentence — what this slide must do]
  Layout: [layout name from asset library]
  Component: [component name from metis-pptx design system, e.g., "Phased Approach", "KPI Summary"]
  Content:
    - [bullet 1]
    - [bullet 2]
  Placeholder: [describe any visual placeholder, or "none"]

SLIDE 2
  ...

Step 10: Iterate by tightening, not reinventing

On revision passes: shorten, sharpen, simplify, improve emphasis, reduce clutter — but preserve the architecture if it is working.

Rebuild from scratch only if the existing structure is fundamentally wrong.

Quality Gates

Standard Quality Bar (Both Modes)

Good output:

• Clearly reflects source materials — nothing invented
• Uses a reusable slide pattern rather than an ad hoc one
• Feels executive and skimmable — key numbers are easy to find
• One job per slide, nothing bleeding into adjacent slides
• Build brief is specific enough for metis-pptx to execute without questions

Weak output:

• Reads like speaker notes
• Mixes multiple slide jobs on one page
• Invents unsupported claims or metrics
• Ignores the outline or existing slide structures
• Over-explains instead of structures
• Produces a vague handoff that requires re-clarification

Completeness Check (Elevation Mode)

Run these four checks before finalizing any elevated output:

Coverage: Does every zone in the client's value stream have at least one capability mapped? If a zone is empty, is that a genuine gap (name it, flag it) or a timing issue (label it "in roadmap")?

Outcome: Does every capability have a stated business outcome? Are outcomes attached to the realization year, not the deployment year? Do individual outcomes add up directionally to the headline number?

Naming: Are all capability names consistent across every slide? Does any name require technical knowledge to understand? Do all names follow the convention from Anchor 3?

Status: Can an executive tell, at a glance, what is live versus building versus aspirational? Is the organization's track record visible — not just its roadmap?

The "So What" Test (Both Modes)

Read the final content as if you are the most senior leader in the room, seeing it for the first time. Ask for every element:

1. "So what?" — Does this create a business implication? If not, cut it or reframe it.
2. "How much?" — Is there a number? If not, is there a directional claim that is still defensible?
3. "By when?" — Is the time horizon clear?
4. "Compared to what?" — Is there a baseline or competitive context?

Any element that fails all four questions belongs in a backup or appendix slide, not the executive version.

Traceability (Both Modes — Critical in Elevation Mode)

Every claim in an executive-facing output must be traceable to a source. This protects the author when a number is challenged, establishes the difference between stated and inferred claims, and ensures source document updates can be reflected systematically.

Claim types — use precisely:

• Actual — A measured result. The strongest claim. Can be stated as fact.
• Target — A stated goal with organizational commitment. Frame as a target, not a result.
• Projection — A modeled figure from assumptions. Carry a qualifier: "projected," "estimated," or "expected."
• Inferred — Derived logically from combining multiple sources. The highest-risk type. Must be validated with a stakeholder before appearing in executive output.

Traceability check (add to Completeness Check):

• Does every number have a source record with document name, location, and claim type?
• Is there at least one "Actual" claim visible — proof that something has been delivered, not just promised?
• Are all "Inferred" claims flagged for stakeholder validation?
• Are all gaps documented, even if they do not appear on the slide?
• If a source document is updated, is it clear which claims need revisiting?

For detailed traceability formats (claim-level, translation-level, gap-level), read references/elevation-reference.md, Section 3.

Common Failure Modes

These appear regardless of client or industry. Watch for them in any mode, but they are especially prevalent when elevating technical content.

Revision Behavior

When the user says "make it punchier," "less salesy," "more executive," "keep the layout, change the wording," or "this belongs in another section":

• Preserve the architecture
• Refine tone and hierarchy
• Tighten wording
• Shift emphasis without losing factual grounding

Writing Style

Concise executive prose: short claims, active voice, low jargon, quantitative where possible, easy to skim. Prefer tight bullets over explanatory paragraphs. Let charts and frameworks carry the argument — use text only to clarify what matters, what changes, and why.