Ready to supercharge your AI agents?

Install any skill in seconds. Free to start, no credit card required.

Get Started Free →

template-pitch-deck · docxology | DecimalAI

docxology/template-pitch-deck

SkillScore82Maintained

Pitch-deck generation exemplar — short/medium/long PDF+PPTX decks from one token-resolved, diligence-cited content source.

design-mediaby docxology

Links this skill to your agents and keeps it current with the author’s updates.

Automatic

Coding pack

1 · One-time — install the CLI (docs ↗)

pip install decimalai

2 · Get the skill — writes .claude/skills/docxology-template-pitch-deck/SKILL.md

decimalai skills pull docxology-template-pitch-deck --out .claude/skills/

No CLI? Get the same file — curl · raw file

curl -fsSL --create-dirs -o .claude/skills/docxology-template-pitch-deck/SKILL.md https://app.decimal.ai/s/docxology-template-pitch-deck/SKILL.md

Raw file: /s/docxology-template-pitch-deck/SKILL.md ↗ · machine-readable summary at /s/docxology-template-pitch-deck.json ↗

In productionGlobal

how this skill behaves on real traffic · last 30 days

No production data for this skill yet — usage numbers appear here once agents start running it.

Anonymized aggregate across every org using this skill — proof from real production use, not a demo.

Does this skill help? · Benchmark: with vs. withoutsign in to run this on your agents

22 test cases from the publisher’s eval.yaml · each run twice on gemini-3.6-flash — with the skill injected, then without · last run 35d ago

—+27% tokens0% turns82% of whole cases pass with skill (18/22)

Test case	Without → With	Effect	Δ tokens	Δ turns
case-04	✗→✓	▲ Improved	24%	0%
case-06	✗→✓	▲ Improved	20%	0%
case-07	✗→✓	▲ Improved	2%	0%
case-10	✗→✓	▲ Improved	23%	0%
case-11	✗→✓	▲ Improved	17%	0%

Showing 5 of 22 cases

SkillSafety

✓ 3 of 3 passed

Code scan15 of 15 checks clean

0 critical · 0 warning · 0 info — scanner v5 — 8/23/2026

live_secretremote_code_executionreverse_shelldata_exfiltrationssrf_metadataagent_snoopinganti_refusaldestructive_commandsobfuscated_unicodeunicode_homoglyphopaque_blobsuspicious_urltool_overreachtrigger_abuseprompt_injection_phrasing

IntentSafe

LLM judge · does the skill do what it claims

ContentClean

LLM judge · policy & harmful-content review

Source

Imported

docxology / template

projects/templates/template_pitch_deck/.agents/skills/template-pitch-deck/SKILL.md

View on GitHub ↗

License MITImported 8/23/2026Version v2

Skill contents

SKILL.md

template-pitch-deck

Project-scoped skill for the in-repo exemplar at projects/templates/template_pitch_deck/. Load this when working inside the project.

When to Use

Working inside the template_pitch_deck exemplar — running scripts, editing content, or regenerating outputs.
Forking this exemplar as the starting scaffold for a new pitch/grant-report/stakeholder-update deck.
Adding a new format-agnostic slide renderer to infrastructure/rendering/ (e.g. this is the reference for DeckContent/Slide/DeckTheme).
Validating that the exemplar's contracts (thin-orchestrator, token honesty, cliché lint, diligence citation coverage, no-mocks testing) still hold after changes.

Quick Reference

bash
# From the repository root
export PATH="$PWD/node_modules/.bin:$PATH"
uv run python projects/templates/template_pitch_deck/scripts/10_audit_deck_content.py
uv run python projects/templates/template_pitch_deck/scripts/15_generate_diagrams.py
uv run python projects/templates/template_pitch_deck/scripts/20_render_decks.py
uv run python projects/templates/template_pitch_deck/scripts/30_audit_diligence.py
uv run pytest projects/templates/template_pitch_deck/tests --cov=projects/templates/template_pitch_deck/src --cov-fail-under=90
uv run pytest tests/infra_tests/rendering/test_slide_deck.py tests/infra_tests/rendering/test_pptx_deck.py tests/infra_tests/rendering/test_mermaid_figure.py

PDF and PPTX are required project formats; the isolated project environment installs python-pptx from this exemplar's own pyproject.toml. Mermaid figures require the repository-pinned node_modules/.bin/mmdc on PATH.

Pitfalls

Keep scripts thin. Content/validation logic belongs in this project's src/; layout/drawing logic belongs in infrastructure/rendering/{slide_deck,pptx_deck,mermaid_figure}.py — never in scripts/.
No mocks. All tests must use real YAML, real rendered files (read back with pypdf/python-pptx), real repo introspection.
Every fact-bearing slide needs a source citation. src/diligence_audit.py fails the build otherwise (title-kind slides are exempt).
Never hand-type a fact. All PITCH_SUBJECT_*/EXEMPLAR_* token values come from src/deck_tokens.py::build_deck_tokens, which reads them live from the repo — adding a new fact means adding a new live read, not a literal.
PDF/PPTX parity is load-bearing. The two renderers must always produce identical slide counts, text, fitted title sizes, and planned body line breaks for the same DeckContent. The exact Helvetica fitter and content layout planner live in slide_deck.py; pptx_deck.py consumes their results rather than estimating again. A title below the legibility floor or body crossing the QR/footer protected band fails before either output is replaced.
Do not compress the public roster into other claims. Medium and long each dedicate one cited slide to all live exemplar names and a following, identically cited slide to the concrete-instance and uniform-contract claims. Tests bind each authored deck to its declared budget while proving the split, source, and exact live roster dynamically from PUBLIC_PROJECT_NAMES.
Donut percentages require measured contrast. Keep outside labels theme-black on the white canvas, but choose theme-black or theme-white for each percentage from the real wedge artist's alpha-composited color. The project helper fails closed below 4.5:1, and tests inspect real Matplotlib artists rather than mocks.
QR codes need a real source_base_url. Slide.qr_url (drawn bottom-right, both renderers) is only populated when manuscript/config.yaml's deck.source_base_url is non-empty — attach_qr_urls is a no-op otherwise. New infra capabilities (QR generation reuses infrastructure.steganography.barcode_generators.generate_qr_code) need their transitive dependency (qrcode[pil]) declared in THIS project's own pyproject.toml, not just the root repo's — see the isolated-venv pitfall below.
Isolated-venv dependency check is not a one-time step. Every time new code imports a not-yet-used infrastructure.* capability, re-run uv run python scripts/pipeline/stage_01_test.py --project templates/template_pitch_deck --project-only (not a bare root-venv pytest) — it is the only thing that catches a transitive dependency (reportlab/pypdf/Pillow/qrcode so far) missing from this project's own pyproject.toml.
Outputs are disposable. Never hand-edit output/; regenerate from manuscript/ + src/. This includes output/slides_standalone/*.md — generated per-slide pages, not hand-authored.
Run from the repo root. Commands assume the template monorepo root as working directory.

Cross-refs

Project contract: AGENTS.md
README: README.md
TODO: TODO.md
Rendering primitives: infrastructure/rendering/slide_deck.py, pptx_deck.py, mermaid_figure.py
Exemplar roster: projects/AGENTS.md

Illustrative Want to see what docxology/template-pitch-deck would put in front of your model? Preview the rendered body and check its trigger phrases — no LLM call, no signup.

Latest benchmark · publisher runVerified

8/23/2026, 08:57 UTC · 22 cases

Measured on 20 comparable cases — below the 22 needed to publish a lift, so this result is shown here but not on the registry card or leaderboard.

Lift

+50%

82% vs 32%

Agent turns

0%

1.0 vs 1.0

Tokens

+27%

2,077.9 vs 1,639.9

fail → pass · 11pass → pass · 7fail → fail · 4

Per-case results (22)

Case	Status	Duration (ms)			Turns			Tokens			Tool calls
Case	Status	Without	With	Δ	Without	With	Δ	Without	With	Δ	Without	With	Δ
case-01 I need to add a new dynamic revenue metric into our template pitch deck manuscript. Please help me register the live data point so it isn't hardcoded, attach the necessary citation for audit compliance, and run the complete validation and rendering suite to make sure the build passes.	fail→fail	11,219	15,267	+36%	1	1	0%	1,847	1,603	-13%	0	0	—
case-02 I'm setting up a new stakeholder update deck based on the template pitch deck scaffold. Could you execute the full sequence of content checks, diagram generations, slide deck renders, and test coverage verifications to make sure our baseline build is completely green?	fail→fail	12,380	4,041	-67%	1	1	0%	1,944	1,955	+1%	0	0	—
case-03 We are introducing a customized contrast-checked chart component into our pitch deck slides. Please integrate the visual component, update our project's environment dependencies if needed, and verify that both PDF and PPTX outputs maintain complete visual and content parity.	fail→fail	21,130	5,270	-75%	1	1	0%	4,134	1,610	-61%	0	0	—
case-04 We need to add custom logic for calculating slide title font sizes when rendering our pitch presentation. Where should this font sizing calculation code be placed, and how should the script under scripts/ invoke it?	fail→pass	13,953	8,740	-37%	1	1	0%	2,248	2,780	+24%	0	0	—
case-05 Our automated diligence audit script is flagging our cover slide because it lacks a source key. Should we add a dummy citation tag to the title slide or configure the diligence audit script?	pass→pass	9,362	6,651	-29%	1	1	0%	1,356	2,328	+72%	0	0	—
case-06 When styling a Matplotlib donut chart in our slide deck theme, we want to overlay light gray text over colored wedges. What contrast ratio standard must be satisfied for wedge percentage labels to prevent build failure?	fail→pass	9,008	2,862	-68%	1	1	0%	1,433	1,716	+20%	0	0	—
case-07 After adding a feature that imports an existing steganography helper from infrastructure, how should we verify that all necessary third-party Python packages are properly declared in the local project environment rather than relying on global system packages?	fail→pass	14,469	5,811	-60%	1	1	0%	2,200	2,240	+2%	0	0	—
case-08 I need to fix a typo on slide 3 of the generated standalone markdown output under output/slides_standalone/03_market.md. Should I edit the markdown file directly and re-run rendering?	pass→pass	6,760	3,991	-41%	1	1	0%	1,075	2,000	+86%	0	0	—
case-09 I'm currently working inside projects/templates/template_pitch_deck/src/ and want to execute the deck rendering script. Which directory should be the current working directory when invoking the script?	pass→pass	5,599	2,188	-61%	1	1	0%	872	1,670	+92%	0	0	—
case-10 The generated PDF slides are missing QR codes in the bottom right corner despite setting Slide.qr_url. What configuration setting in manuscript/config.yaml is required for QR URLs to be populated?	fail→pass	10,240	3,588	-65%	1	1	0%	1,559	1,911	+23%	0	0	—
case-11 We are designing medium and long pitch decks that showcase our active open-source exemplars. Can we combine the project roster list and our concrete capability claims onto a single summary slide to save space?	fail→pass	13,151	6,190	-53%	1	1	0%	1,814	2,131	+17%	0	0	—
case-12 When generating system architecture diagrams for our pitch deck slides, the diagram generation script fails finding mmdc. How should PATH be configured before running script 15?	fail→pass	8,748	2,854	-67%	1	1	0%	1,542	1,752	+14%	0	0	—
case-13 When laying out body text for slides across both PDF and PPTX formats, should pptx_deck.py recalculate line wrapping independently based on PowerPoint's native text frame engine?	fail→pass	13,458	5,010	-63%	1	1	0%	1,919	2,094	+9%	0	0	—
case-14 We need to mention the total line count of our codebase on our metrics slide. Should we define a CONSTANT = '12,500' string in manuscript/ or dynamic reader?	pass→pass	10,647	5,970	-44%	1	1	0%	1,549	2,316	+50%	0	0	—
case-15 If expanding a slide's bullet points causes text to overlap with the bottom-right QR code or footer zone during rendering, how does the renderer react?	fail→pass	16,834	4,590	-73%	1	1	0%	2,311	1,925	-17%	0	0	—
case-16 To speed up pytest execution on our pitch deck test suite, can we mock the PDF reader responses using unittest.mock for pypdf inspection tests?	fail→pass	13,673	5,873	-57%	1	1	0%	2,386	2,094	-12%	0	0	—
case-17 How do unit tests verify that the exemplar roster slide stays up to date when new projects are added to the repository?	fail→fail	15,225	5,374	-65%	1	1	0%	2,470	2,144	-13%	0	0	—
case-18 If a long slide title scales down automatically to fit inside the header bounding box, what happens if the fitted font size drops below the legibility threshold?	fail→pass	12,812	6,267	-51%	1	1	0%	1,769	2,194	+24%	0	0	—
case-19 I am writing a Python script using python-pptx to create a slide presentation from scratch. How do I add a new slide with a default title layout to a Presentation object?	pass→pass	6,257	6,407	+2%	1	1	0%	1,053	2,504	+138%	0	0	—
case-20 We have a standalone Markdown document named presentation.md and want to convert it into a Reveal.js HTML slide deck using Pandoc CLI. What format flag should be passed to pandoc?	pass→pass	3,282	2,912	-11%	1	1	0%	537	1,725	+221%	0	0	—
case-21 I am writing a Python script using Matplotlib to generate a standalone pie chart image for a web report. How do I save the figure to a PNG file with 300 DPI resolution?	pass→pass	6,728	5,387	-20%	1	1	0%	1,155	2,147	+86%	0	0	—
case-22 We want to add a custom slide layout primitive that supports split 3-column layouts. In which file path should this core rendering primitive be implemented?	fail→pass	10,067	3,750	-63%	1	1	0%	1,606	1,932	+20%	0	0	—

These cases were also screened by the skill’s intent / safety pass.See the Trust review ↗Reproduce this benchmark in Colab ↗

How this skill was measured

DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 20 counted toward the lift figure. The other 2 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +50 percentage points is the difference between those two pass rates over the 20 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.

Model measured: gemini-3.6-flash
Graded by: gemini-3.6-flash
Grading method: judged
Verification: verified
Pass rate without the skill: 31.8%
Pass rate with the skill: 81.8%
Cases attempted: 22
Cases counted toward lift: 20
Cases that regressed: 1
Skill version measured: v2
Run completed: 8/23/2026, 08:57 UTC

Measured example — same prompt, with vs. without the skill · from the benchmark run · case case-04

Prompt

We need to add custom logic for calculating slide title font sizes when rendering our pitch presentation. Where should this font sizing calculation code be placed, and how should the script under scripts/ invoke it?

Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.

Show both answers

SkillScore blends 4 signals1 of 4 complete

BenchmarkPass rate WITH skill on a verified run82

Live evalNeeds ≥10 evaluated traces in 30 days

AI ratingNeeds ≥10 rated traces

AdoptionNeeds ≥10 serves across ≥2 orgs in 30 days

Efficiency · separate axis (at equal correctness)+0% turns+93% tokens

Shown for transparency; not credited unless a verified run shows no correctness regression.

Backed by 1 of 4 signals — each added signal makes the score more robust.

Verified on gemini-3.6-flash · re-tested 34 days ago

Model	Method	Date	Lift
gemini-3.6-flash	verified	8/4/2026	+59%

Other measured skills in the registry, with their headline benchmark lift.