Create Demo PPTX

Create Datadog Demo PPTX

Turn customer meeting notes into a polished, Datadog-branded demo deck built on the bundled template assets/2026_Datadog_Presentation_Template.pptx (~20 pre-designed slides, purple/violet/white palette, Datadog logo). Input: the user's meeting notes. Output: a finished .pptx following the preferred slide flow.

Setup

Set these paths (relative to this skill's directory) before running commands:

SCRIPTS="$(pwd)/scripts"          # or the absolute path to this skill's scripts/
TEMPLATE="$(pwd)/assets/2026_Datadog_Presentation_Template.pptx"

Workflow

Copy this checklist and track progress:

- [ ] Step 1: Ask clarifying questions
- [ ] Step 2: Build the slide plan (markdown table)
- [ ] Step 3: Generate any needed images (mock-ups / diagrams)
- [ ] Step 4: Build the deck (unpack → duplicate → edit → clean → pack)
- [ ] Step 5: Visual Check Loop (mandatory) → fix → re-verify
- [ ] Step 6: Generate the architecture diagram (invoke create-architecture-excalidraw)

Step 1 — Ask clarifying questions first

Before creating the pptx, ask the user for anything missing from their notes. Do not fabricate content. At minimum confirm:

• Customer name, presenter name(s), and contact details (slides 1 and 19)
• Presentation date and delivery context
• The top 3 customer challenges (slide 3)
• The 3 Datadog use cases and which maps to each challenge (slides 4, 6-11)
• Current state vs. target state details (slides 5, 13)
• Competitors / incumbents for the capability comparison (slide 14)
• Business value / ROI figures (slide 16)
• A relevant customer story and strategic + technical next steps (slide 18)
• Whether to include the optional reference resources slide (slide 20)
• Which use cases need a screenshot mock-up, architecture, or system diagram
• Output file path/name for the finished deck

Use the AskQuestion tool when available; otherwise ask conversationally. Ask only what is genuinely missing — infer the rest from the notes.

Step 2 — Build the slide plan

Write the plan as a markdown table before touching any files. Map each row of the preferred flow to the template source slide it will be duplicated from. See references/slide-flow.md for the full 20-slide flow and mapping guidance. Analyze the template first:

python3 $SCRIPTS/thumbnail.py "$TEMPLATE"
python -m markitdown "$TEMPLATE"

Step 3 — Generate images

Before editing slides, generate images for any slide that needs a screenshot mock-up, architecture diagram, or system diagram (typically the use-case detail slides 7, 9, 11).

Step 4 — Build the deck

Unpack, duplicate the mapped source slides, reorder in <p:sldIdLst>, edit content, clean, and pack. Complete all structural changes before editing text. Full commands, editing rules, and pitfalls are in references/editing-and-qa.md.

python3 $SCRIPTS/office/unpack.py "$TEMPLATE" unpacked/
python3 $SCRIPTS/add_slide.py unpacked/ slide7.xml   # duplicate a source slide
# edit unpacked/ppt/slides/slide{N}.xml with StrReplace
python3 $SCRIPTS/clean.py unpacked/
python3 $SCRIPTS/office/pack.py unpacked/ output.pptx --original "$TEMPLATE"

Preserve Datadog branding — only replace placeholder text and visuals; never alter the palette, logo, or master styling.

Step 5 — Visual Check Loop (mandatory)

After packing, render every slide to an image and review it. Font size problems, text overflow, and truncation are invisible until rendered — this step is not optional.

1. Render to images (soffice → pdftoppm), then Read every slide image.
2. Build a fix plan table of issues before fixing anything.
3. Fix fonts / content in the slide XML.
4. Re-render and re-verify affected slides. One fix often creates another.
5. Repeat until a full pass reveals no new issues.

Detailed issue checklist, font-size table, and fix snippets are in references/editing-and-qa.md.

Step 6 — Generate the architecture diagram

The deck isn't the only thing these discovery notes are good for. Once the deck passes Step 5, the same meeting notes almost always carry enough technical detail (environments, services, compliance, network constraints) to also sketch the customer's monitoring architecture — so this is a mandatory final step, not an optional add-on.

Invoke the create-architecture-excalidraw skill via the Skill tool, passing it the original discovery/meeting notes from Step 1 as-is. Don't pre-filter or summarize them — that skill has its own clarifying-questions step and will ask about anything it still needs (intake region, compliance requirements, collection-method preferences) that the notes don't cover.

The end result of one request is both deliverables: the finished .pptx and an Excalidraw architecture diagram (or its .excalidraw JSON fallback) built from the same notes.

Resources

• references/slide-flow.md — the preferred 20-slide flow and template mapping
• references/editing-and-qa.md — build commands, font tables, fix snippets, visual QA loop
• scripts/ — PPTX editing utilities (unpack.py, add_slide.py, clean.py, pack.py, thumbnail.py, office/soffice.py)
• assets/2026_Datadog_Presentation_Template.pptx — the branded source template