insight-storyteller
Skill2 starsChanged 36 days agoMITView on GitHub
Turns analysis results into a decision-ready story: a one-line answer, a short executive summary, a slide-by-slide outline with the right chart for each point, and stakeholder messages for Slack, email, and leadership. Use this skill when the user says "write up these findings", "make a readout", "turn this into a deck/story", "executive summary", "how do I present this", "draft the update for leadership", "experiment readout", or shares analysis output that needs to be communicated. Final stage of the Product Analytics pack.
In this repository21 other files it gives its agents.
- analytics-question-framingSkill
- clinical-ai-evaluation-designSkill
- cohort-retentionSkill
- design-systemSkill
- engineering-plannerSkill
- experiment-designSkill
- frontend-setupSkill
- gxp-part11-checkerSkill
- implementation-specsSkill
- market-researchSkill
- metric-root-causeSkill
- opportunity-finderSkill
- opportunity-sizingSkill
- prd-evaluatorSkill
- prd-writerSkill
- product-strategy-canvasSkill
- security-foundationSkill
- strategy-sprintSkill
- trial-eligibility-matcherSkill
- trial-failure-investigatorSkill
- user-researchSkill
The file
We split this skill into the rules it gives an agent
5 rules. About 0 are common, 5 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈5
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 21
Before writing anything, complete this sentence in under 25 words:
- Distinctive: line 25
If you can't, the analysis isn't finished.
- Distinctive: line 29
Use this order (the Minto "SCR" structure, with the answer first):
- Distinctive: line 47
Every chart gets an action title that states the takeaway ("Android drove 80% of the activation drop"), not a label ("Activation by …
- Distinctive: line 97
If the user has slide tooling available (for example a PowerPoint or Slides skill), offer to build the deck from the outline.
≈ “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
Product Analytics pack · Stage 5.
analytics-question-framing→metric-root-cause/cohort-retention→opportunity-sizing→experiment-design→insight-storyteller
Insight Storyteller
Stakeholders don't need everything you found. They need the answer, why they should believe it, and what to do. Lead with that; put the method in the appendix.
Step 1 — Find the one thing
Before writing anything, complete this sentence in under 25 words:
"We found
<insight>, which means<implication>, so we recommend<action>."
If you can't, the analysis isn't finished. Say which question is still open instead of padding the story.
Step 2 — Structure: Situation → Complication → Resolution
Use this order (the Minto "SCR" structure, with the answer first):
- Answer first: the one-line insight and recommendation.
- Situation: what the audience already knows and agrees with (the baseline, the goal).
- Complication: what changed or what's at risk. This is why it matters now.
- Resolution: the evidence (3 points maximum), the recommendation, its expected impact, and the ask (decision, budget, owner, date).
Step 3 — Match each point to a chart
| The point is about… | Use | Avoid |
| Change over time | Line chart, with the event annotated | Pie charts, 3D |
| Comparing groups | Sorted horizontal bar chart | Unsorted bars, too many colors |
| Part of a whole changing | Stacked bar (≤ 5 parts) or waterfall | Multiple pies |
| What drove a change | Waterfall (mix vs rate, or by segment) | Long tables |
| Retention | Cohort curves or a heat-mapped triangle | Single averages |
| Experiment result | Point estimate with confidence interval vs zero | Bars without error ranges |
Every chart gets an action title that states the takeaway ("Android drove 80% of the activation drop"), not a label ("Activation by platform"). Highlight the one series that matters and grey out the rest.
Step 4 — Calibrate confidence honestly
State how sure you are and why: sample size, data quality, causal versus correlational evidence. Use plain words: "We're confident…", "The evidence suggests…", "We can't yet tell whether…". Never overstate a correlation as a cause.
Step 5 — Write for each audience
Produce what the user needs from this list:
- Executive summary (5 lines): answer, why, impact in numbers, recommendation, ask.
- Slide outline (6–10 slides): title slide with the answer → situation → complication → 2–3 evidence slides → recommendation and impact → next steps and owners → appendix (method, definitions, caveats).
- Slack update (3–4 lines, plus a link to details).
- Email to leadership (subject line that states the finding, 1 short paragraph, bullets, the ask).
- Speaker notes for each slide (what to say, the question you'll likely get, and the answer).
Output template
# <Action title: the insight in one line>
**Recommendation:** <action> · **Impact:** <number, range> · **Ask:** <decision needed, by when>
**Confidence:** <High/Medium/Low>, because <reason>
## Executive summary
1. …
2. …
## Slide outline
| # | Action title | Chart | Key number | Speaker note |
|---|---|---|---|---|
## Messages
**Slack:** …
**Email subject:** …
**Email body:** …
## Appendix
- Method and definitions
- Caveats and what we couldn't testQuality check before handing over
- Could a busy executive get the point from the title and first line alone?
- Does every slide title state a takeaway?
- Is every number traceable to the analysis, with units and time period?
- Are caveats present but not burying the message?
- Is there a clear owner and date for the next step?
If the user has slide tooling available (for example a PowerPoint or Slides skill), offer to build the deck from the outline.
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.

