JPeetz/agent-skills / skills/documentation-content/presentation-prompt-pipeline
presentation-prompt-pipeline
VerifiedUse this skill when building or repairing a presentation with an AI model — the six staged prompts (blueprint → structure → story frame → visual direction → content → build) that turn an idea into a deck with one message, each stage producing an artifact and a checkable gate, with prompt templates, eight named failure modes, and a deck_lint.py script that mechanically enforces short titles, tight bullets, speaker notes, a two-font and three-colour ceiling.
| File | Kind | Size |
|---|---|---|
| CHANGELOG.md | markdown | 1013 B |
| LICENSE | other | 1.0 KB |
| SKILL.md | skill md | 8.1 KB |
| evals/evals.json | config | 6.4 KB |
| references/six-prompts-cheatsheet.md | markdown | 1.6 KB |
| scripts/deck_lint.py | script | 7.6 KB |
evals/evals.json
View on GitHub{
"skill_name": "presentation-prompt-pipeline",
"evals": [
{
"id": 1,
"prompt": "I need a 20-slide deck for our Q3 business review next week. Can you write it?",
"expected_output": "Do not jump straight to slide copy. Start at stage 1 (blueprint): ask for or derive the ONE message as a falsifiable claim (not 'Q3 review'), the audience and what they care about, 2-3 emotional angles, and only then challenge the 20-slide count ('20 slides for one meeting usually means three decks'). Proceed through the stages in order — blueprint, structure, story frame, visual direction, content, build — each in its own prompt with the previous artifact as input, and state the gate that must pass before the next stage.",
"expectations": [
"Refuses to produce slide copy immediately and starts with the blueprint stage",
"Requires the single message to be a claim, not a topic",
"Questions or justifies the slide count instead of accepting it",
"Keeps the stages in order with the previous artifact as input",
"Names a gate/check before advancing"
],
"should_trigger": true
},
{
"id": 2,
"prompt": "My sales deck has 14 slides and every one is a title plus six bullet points of 20 words. It reads like a document. What do I change?",
"expected_output": "Diagnose it as bullet walls, not a formatting issue: the deck skipped the blueprint/structure stages, so each slide has no purpose line and the copy is carrying the argument. Fix in order: (1) write the one-sentence message and each slide's one-line purpose, deleting slides without a distinct job; (2) map the remainder onto Hook/Problem/Insight/Solution/Proof/Takeaway with a single ask at the end; (3) rewrite titles to <=6 words and bullets to 3-5 of <=10 words, moving the prose into speaker notes so the argument is spoken, not displayed; (4) run deck_lint.py to enforce the limits mechanically and render the slides to check overflow.",
"expectations": [
"Diagnoses missing blueprint/structure rather than treating it as a styling problem",
"Requires one-line purpose per slide and deletion of slides without a distinct job",
"Applies the story frame with exactly one action on the final slide",
"Enforces <=6-word titles and 3-5 bullets of <=10 words, moving prose into notes",
"Recommends deck_lint.py plus a rendered visual check"
],
"should_trigger": true
},
{
"id": 3,
"prompt": "Which chart should I use for our revenue by region over 8 quarters, and how do I phrase the slide title?",
"expected_output": "Stage 4 answer: bar for the regional comparison (or line for trend within each region if the change over time is the point), one chart per idea, one number called out. Stage 5 answer for the title: a <=6-word claim, not a label — e.g. 'EMEA grew, APAC stalled' rather than 'Revenue by region Q1-Q8' — with the takeaway in the speaker note and at most a three-colour palette. Note the deck-level constraint: the plot must support the deck's single message.",
"expectations": [
"Chooses the chart from the communicative job (comparison vs trend) and justifies it",
"Rejects label-style titles in favour of a <=6-word claim",
"Keeps to one chart/one idea/one number",
"Ties the choice back to the deck's single message"
],
"should_trigger": true
},
{
"id": 4,
"prompt": "Can you check this deck.pptx for me — titles too long, too many bullets, missing speaker notes?",
"expected_output": "Run the bundled linter rather than eyeballing it: `uv run --with python-pptx scripts/deck_lint.py deck.pptx` (or with tuned --max-title-words/--max-bullets flags), report its JSON findings per slide, and follow up on what the linter cannot see — render the slides and check overflow, contrast and orphaned bullets. Do not claim the deck is fine without running something.",
"expectations": [
"Uses the deck_lint.py script instead of manual counting",
"Reports concrete per-slide findings with the rule that fired",
"Adds the visual/render check the linter cannot perform",
"Does not assert quality without an actual run"
],
"should_trigger": true
},
{
"id": 5,
"prompt": "Give me the visual direction for a 6-slide investor pitch about a grid-scale battery startup.",
"expected_output": "Stay in stage 4 and output per-slide: the emotion the visual carries, the chart or single big number for the data, icon/grid treatment for lists, one three-colour palette with hex codes, and one heading/body font pairing available in standard Office/Google fonts. Describe composition, not stock-photo subjects, and keep the palette/fonts constant across all six slides.",
"expectations": [
"Produces per-slide emotional intent plus chart-choice reasoning",
"Supplies one three-colour palette with hex codes and one font pairing",
"Describes composition instead of naming stock photos",
"Keeps the visual system constant across slides"
],
"should_trigger": true
},
{
"id": 6,
"prompt": "Convert this deck.pptx to PDF and rename the file for me.",
"expected_output": "Out of scope for this skill — it is a format/mechanical operation (soffice/LibreOffice conversion), not deck planning or review. Point at the format-conversion route and do not run the six-stage pipeline.",
"expectations": [
"Recognises the request as format conversion, outside the pipeline",
"Does not invoke the blueprint/structure/content stages",
"Points at the mechanical conversion route instead"
],
"should_trigger": false
},
{
"id": 7,
"prompt": "Write me a 40-slide internal wiki page about our onboarding process.",
"expected_output": "Not a presentation task: the artifact requested is a reference document, where length and completeness are assets. Say so and suggest a document format rather than applying slide constraints (six-word titles, five bullets) that would destroy the reference value. If a slide version is genuinely wanted later, the pipeline applies to a summary deck derived from the document.",
"expectations": [
"Identifies the artifact as a document, not a deck",
"Declines to impose slide constraints on reference material",
"Offers the pipeline only for a derived summary deck"
],
"should_trigger": false
}
]
}
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| writing-skillsobra/superpowersUse when creating new skills, editing existing skills, or verifying skills work before deployment | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| xlsxanthropics/skillsUse this skill any time a spreadsheet file is the primary input or output. This means any task where the user wants to: open, read, edit, or fix an existing .xlsx, .xlsm, .xltx, .csv, or .tsv file (e.g., adding columns, computing formulas, formatting, charting, cleaning messy data); create a new spreadsheet from scratch or from other data sources; or convert between tabular file formats. Trigger especially when the user references a spreadsheet file by name or path — even casually (like "the xlsx in my downloads") — and wants something done to it or produced from it. Also trigger for cleaning or restructuring messy tabular data files (malformed rows, misplaced headers, junk data) into proper spreadsheets. The deliverable must be a spreadsheet file. Do NOT trigger when the primary deliverable is a Word document, HTML report, standalone Python script, database pipeline, or Google Sheets API integration, even if tabular data is involved. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| shipping-and-launchaddyosmani/agent-skillsPrepares production launches. Use when preparing to deploy to production, or when asking what needs to be in place before shipping. Use when you need a pre-launch checklist, when setting up monitoring, when planning a staged rollout, or when you need a rollback strategy. | Stars: 119k | Updated | Claude CodeCodex | Verified | Install |
| childhoodthedotmack/claude-memBabysit a shipped release after merge, publication, or deployment. Verify published packages, production behavior, installed plugins, and late CI or user reports until the observation window is clean; use babysit for PR readiness before release. | Stars: 118k | Updated | Claude CodeCodex | Verified | Install |