pitchdeck-review
Skill0 starsChanged 50 days agoMITView on GitHub
Use after pitchdeck-build produces a deck that needs judging — runs QA, reviews rendered slides against the four rules (storyline, visual-over-words, insight, font size), and emits scope-tagged findings for the loop-back. Trigger on "pitchdeck-review", "review the deck", "is this deck good", or after build handoff option 1. Do NOT use as a substitute for build QA (build owns its own gates) — this is the independent post-build judgment.
In this repository9 other files it gives its agents.
- llms.txtllms.txt
- assets-generatorSkill
- audience-fitSkill
- brand-designSkill
- humanizerSkill
- pitchdeck-buildSkill
- pitchdeck-grindingSkill
- pitchdeck-pitch-meSkill
- pptxSkill
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 1 is distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 24
No deck file yet → pitchdeck-build
- Distinctive: line 58
Read every JPG. Judge against the four rules + UX:
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 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
Review — judge the deck, emit scoped findings
Independent post-build review: run both QA halves, judge the rendered slides against
the four rules, and produce findings that carry scope so the loop sends each defect
to the skill that owns it.
Core principle: A review without a scoped next step is a complaint. Every finding
names its scope and its fix target; the pipeline's loop (max 2 auto-iterations, then
REVIEW_BLOCKED) only works if scopes are honest.
When to Use
pitchdeck-buildfinished (its own gates passed)- User asks "review the deck" / "is this ready"
- Re-review after a loop-back iteration (bounded by
review.maxAutoIterations)
Do NOT use:
- Build QA hasn't run — build fails first, review shouldn't double-report it
- No deck file yet →
pitchdeck-build
Hard Gates
GATE 1: BOTH QA halves run before any judgment:
deckto qa static + deckto qa render (when tools present).
Render missing → say so; do not review from XML alone and call it visual.
GATE 2: EVERY finding carries scope: storyline | assets | deck.
Scope decides the loop-back target — missing scope = unusable finding.
GATE 3: EVERY finding carries rule + evidence + fix.
"Looks meh" without the slide number, the rule, and the fix is not a finding.
GATE 4: Findings are written to deck/<slug>/qa/findings.json in the qa report
schema, then run `deckto qa report` for the loop directive.
Honor REVIEW_BLOCKED — after maxAutoIterations, stop looping and ask the user.Phase 1 — Mechanical evidence
npx deckto qa static deck/<slug>/deck.pptx --json
npx deckto qa render deck/<slug>/deck.pptx --out deck/<slug>/qa/render
npx deckto qa storyline deck/<slug>/storyline.md --json # cross-check intent vs outputStatic failures (font floor, word budget, insight, visual, arc) are already findings —
import them with their scope (deck for build defects; if the same root cause lives in
storyline.md, scope it storyline).
Phase 2 — Visual review of rendered slides
Read every JPG. Judge against the four rules + UX:
| Rule | What to look for | Typical scope |
| 1. Storyline | does the deck have a spine visible in the slide order? master title claim actually argued? | storyline |
| 2. Visual > wordy | paragraph walls, text-through-images, tiny-but-legible? | deck (build) or storyline (if body already over budget in source) |
| 3. Insight | headline carries a "so what"? or is it a label pretending? | storyline if source insight is weak, deck if notes lost it |
| 4. Font size | readable at back of room (judge from the image, not XML) | deck |
| UX | overlaps, off-center, contrast failures, inconsistent motif, placeholder artifacts | deck or assets |
Judge the image, not the intent. The storyline's intention doesn't reach the audience — only pixels do.
Phase 3 — Findings + loop directive
Write deck/<slug>/qa/findings.json:
{ "file": "deck/<slug>/deck.pptx",
"findings": [
{ "scope": "assets", "rule": "visual-quality",
"slide": 7, "evidence": "flow.png rendered with overlapping node labels",
"fix": "regenerate flow.svg with wider nodes, re-run assets render" }
] }Then:
npx deckto qa report --findings deck/<slug>/qa/findings.json --iterations <n>Output tells you the loop action: which scopes to re-invoke (storyline →
pitchdeck-grinding, assets → assets-generator, deck → pitchdeck-build) or
REVIEW_BLOCKED → present remaining findings to the user and stop.
Phase 4 — Verdict
Present: pass/fail per rule, the scoped findings table, the loop directive, and the three options (fix via loop / accept with noted risks / stop). Honesty over smoothness: if the deck is thin on insight, say so with the slide numbers — a soft review costs the user their pitch.
Common Mistakes
| Mistake | Consequence |
| Reviewing only the XML | visual defects (overlaps, contrast) invisible |
| Findings without scope | loop can't route; iteration stalls |
| Re-scoping build defects as storyline | grinding re-runs unnecessarily; 4× cost |
| Ignoring REVIEW_BLOCKED cap | infinite loop; user never consulted |
| Reviewing intent ("the story is there") instead of pixels | rule 2 and UX pass on paper, fail on screen |
Honesty note
The four-rule judgment is human-design-principled ([A]/[B] sources recorded in
../../docs/research/frameworks-methodology.md); the review applies them as structured
checklists, not as a validated scoring model. Two passes of the same review will not
produce identical numbers — findings with slide-level evidence remain actionable even
when verdicts are judgment calls.
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.

