deck-architect
Skill1 starsChanged 3 months agoMITView on GitHub
Plan the narrative, audience journey, slide sequence, evidence, and decision structure for a presentation before implementation. Use for new decks, major rewrites, executive briefs, technical talks, and source-to-slides conversion.
In this repository17 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- claude-code-styleSkill
- create-deckSkill
- deck-architectSkill
- deck-reviewerSkill
- diagram-designSkill
- review-deckSkill
- speaker-notesSkill
- visual-directorSkill
- claude-code-styleSkill
- create-deckSkill
- deck-reviewerSkill
- diagram-designSkill
- review-deckSkill
- speaker-notesSkill
- visual-directorSkill
The file
We split this skill into the rules it gives an agent
8 rules. About 0 are common, 8 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈8
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 8
Read the storytelling and slide-planning system, resolved relative to this skill directory.
- Distinctive: line 25
Ask only when a missing constraint materially changes the story.
- Distinctive: line 29
When the request explicitly asks to explore, or multiple angles would create materially different decks, compare exactly three concise …
- Distinctive: line 31
Do not write multiple full outlines.
- Distinctive: line 35
Read every source the request depends on.
- Distinctive: line 54
Run a section-continuity pass:
- Distinctive: line 84
Every page has an internal audience takeaway, but a visible takeaway callout is default-off.
- Distinctive: line 99
Do not implement HTML, CSS, Markdown slides, PptxGenJS, PowerPoint, or visual assets.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Plan the deck before it is built
Read the storytelling and slide-planning system, resolved relative to this skill directory. Produce a narrative and slide blueprint, not slide implementation code.
For a non-trivial deck, write or update a deck-local deck-plan.md. When no workspace path exists, return the same plan in the response so the orchestrating workflow can persist it before authoring slides.
1. Build the communication contract
Identify:
- audience, context, and current knowledge
- the decision, belief, or action the presentation should create
- duration and implied page budget
- one-sentence audience promise
- central thesis
- one-sentence Direction Lock
- strongest evidence and weakest unsupported assumption
- language, brand, confidentiality, accessibility, and format constraints
Ask only when a missing constraint materially changes the story. Otherwise record the assumption and continue.
2. Resolve narrative direction
When the request explicitly asks to explore, or multiple angles would create materially different decks, compare exactly three concise narrative forks. For each, state the thesis, best-fit audience or use case, high-level flow, strongest evidence, and main trade-off.
Do not write multiple full outlines. Select or synthesize one Direction Lock before continuing. Skip this step when the direction is already clear.
3. Inspect and classify evidence
Read every source the request depends on. Classify important claims as:
- Verified
- Derived
- Assumption
- Proposal
Record source, date, unit, what the evidence proves, and any limitation or conflict. Never invent numbers, benchmarks, dates, outcomes, systems, customers, quotations, or citations.
4. Design the story spine
Choose a defensible sequence rather than preserving source order. A useful default is:
Reality → Tension → Reframe → Mechanism → Proof → DecisionAdapt it to the deck type. Group pages into small sections with a clear throughline. Every page must have a unique informational job.
Run a section-continuity pass:
- Does each page answer or advance the previous page?
- Is evidence shown close to the claim it supports?
- Does the deck jump from symptom to solution without root cause or mechanism?
- Are conclusions shown before their support?
- Are two pages repeating the same point?
Record a compact Narrative Review. Name only unresolved slide pairs, the gap, and the least disruptive adjustment.
5. Create the slide blueprint
For every page include:
- slide id and section
- role and purpose
- claim-style title
- optional supporting sentence
- internal audience takeaway
- evidence status, source, and source assets
- information shape
- visual form
- layout intent: dominant element, desired density, and intended eye path
- visible content
- speaker-note content
- transition in and transition out
- editing constraints
The informationShape and layoutIntent are semantic requirements. The visual director resolves the final archetype, geometry, and medium from the layout system.
Every page has an internal audience takeaway, but a visible takeaway callout is default-off. Add one only when it contributes a source-grounded implication not already expressed by the title and working area.
Deliverable
Return or write one deck-plan.md containing:
- Presentation Contract
- Evidence Ledger
- Narrative Strategy and Direction Lock
- Story Spine and section map
- Narrative Review
- Page-by-page Slide Blueprint
- Content to remove or move into notes
- Risks, assumptions, missing evidence, and source conflicts
Do not implement HTML, CSS, Markdown slides, PptxGenJS, PowerPoint, or visual assets. Do not propose decoration without a communication purpose. Prefer a shorter coherent deck over a long comprehensive one.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

