the skill

Action Titles

Rewrite slide titles into McKinsey-style action titles — complete "so-what" sentences that state the takeaway, verb-led and quantified, two lines maximum, parallel across a section and laddering to the governing thought. Use whenever the user is working on PowerPoint or deck headlines, slide titles, headings or subheadings; wants titles made action-oriented or "so-what"; asks to fix label-style or merely descriptive titles; or hands over a .pptx, an outline, or slide body content that needs strong titles — even if they don't say "action title". Produces a before-to-after table with rationale, not an edited file.

Turn weak slide titles into action titles: a single declarative sentence that tells the reader the conclusion, so the deck can be read by titles alone. This skill rewrites and explains; it does not edit the source file.

What an action title is

An action title states the so-what, not the subject. It is the answer the slide proves, written as a complete sentence.

The standard this skill enforces:

  • Complete sentence, stating the takeaway. "Enterprise expansion drove 18% revenue growth," not "Revenue" or "Revenue overview."
  • So-what, not description. Say what the data means for the decision, not what the chart shows. Not "Trends in customer churn" but "Churn is concentrated in month-1 customers, so onboarding is the highest-leverage fix."
  • Verb-led and active. A working verb carries the claim. Prefer active voice and present or simple past tense.
  • Quantified where the data allows. Put the number in the title — magnitude, delta, share, or rank — when the body supports it. A specific figure beats "significantly."
  • Two lines maximum, one preferred. Rule of thumb: under ~14 words / ~85 characters. The real test is whether it wraps past two lines at the deck's title size; when editing a real .pptx, check against the actual title width.
  • One message per title. If a title contains "and" joining two distinct claims, it is two slides or a title plus a subtitle.
  • Parallel across a section. Titles within a section share grammatical shape so they read as a list when skimmed.
  • Ladders to the governing thought. Each section's titles should sum to the section's argument, and sections should sum to the deck's single governing thought (Minto pyramid). Flag titles that don't support the level above them.

Workflow

The skill accepts three inputs. Detect which one you have and proceed.

1. A .pptx file

Read the title text per slide without modifying the file:

from pptx import Presentation
prs = Presentation(path)
for i, slide in enumerate(prs.slides, 1):
    title = slide.shapes.title
    text = title.text if title is not None else None
    # If no title placeholder, fall back to the topmost text box by .top

Some decks put the action title in a free text box rather than the title placeholder; if slide.shapes.title is None, take the highest text frame on the slide. Capture the slide's body/chart context too — you need the so-what, and the current title may not contain it.

2. Pasted text or an outline

Treat each heading (or each bullet block that maps to a slide) as one slide. Use the supporting bullets as the evidence for the so-what.

3. Slide body with no title

Infer the takeaway from the content and write the title from scratch.

Rewrite procedure (every input)

  1. Identify the slide's single message — what does this slide prove? If you can't find it in the body, say so rather than inventing a claim.
  2. Draft a verb-led sentence stating that message.
  3. Quantify it from the body if a figure is available; never fabricate a number.
  4. Trim to two lines / ~14 words; cut hedges ("may," "could help"), filler ("overview of," "key"), and label nouns.
  5. Check the title against its section: parallel shape with sibling titles, and does it support the section's claim and the governing thought?

Output

Return a before → after table with a rationale column, grouped by section. Do not edit or return a modified file.

After the table, add a short section-coherence note: state each section's governing thought in one line, confirm the rewritten titles ladder up to it, and flag any title that doesn't fit or any section whose titles don't sum to its claim. If a slide's body lacks a defensible takeaway, list it as "needs a sharper message" rather than forcing a title.

Common failure modes to catch

  • Label titles — a noun phrase ("Market landscape," "Pricing"). Always rewrite to a sentence.
  • Description, not implication — "Chart shows declining margins" states what's visible, not why it matters.
  • Two messages in one — split, or demote the second to a subtitle.
  • Unquantified vagueness — "strong growth," "significant opportunity." Add the figure or cut the adjective.
  • Passive / verbless — "Costs were analyzed." State the finding.
  • Title that fights the chart — the so-what must be the thing the exhibit actually proves.
  • Orphan title — doesn't support the section's argument; either the title or the slide's placement is wrong.

Worked examples

  • "Customer segmentation" → "Three segments hold 80% of profit pool, so targeting should narrow to them"
  • "Competitive landscape overview" → "No incumbent serves the mid-market, leaving an open entry point"
  • "Results of the pricing analysis" → "A usage-based tier lifts ARPU 22% without raising churn"
  • "Implementation considerations and risks" → "Launch is feasible in two quarters; vendor onboarding is the critical-path risk" (note: borderline two-message — consider splitting)
  • "Q3 was a good quarter" → "Q3 revenue rose 14% as net retention hit 118%"