Workplace Writing skill 04 of 13

Reports SKILL.md

The Reports skill makes Claude, Codex or another AI agent write reports that get read and used. It starts from the question the report answers, leads with key findings and what they mean, supports each finding with evidence and charts whose titles state the point, separates facts from interpretation, states limitations and confidence and ends with recommendations that have owners and dates.

reports/SKILL.md, version 1.0.0, 1,493 words.

Reports are written to inform decisions, yet many are read by nobody. They follow the order the work was done (method, data, then findings), bury the conclusion on page twelve, use charts without saying what they show and leave readers to work out what to do. Senior readers in particular look for the answer first and the evidence only if they doubt it.

This skill writes analytical, research, performance and project reports that lead with findings and implications, support them with clear evidence and charts that state their point, separate facts from interpretation, show limitations openly and end with recommendations and next steps. It suits monthly performance reports, research findings, evaluations, audits and end-of-project reports.

Operating principles

  1. Answer first. Key findings and what they mean come in the first page.
  2. Every chart has a headline. The chart title states the point, not just the metric.
  3. Facts, interpretation and recommendation are separate. Readers can see which is which.
  4. Show limitations. Data gaps, sample sizes and assumptions are stated, not hidden.
  5. End with action. Recommendations with owners and dates.

Report types

TypeLead with
Performance report (monthly or quarterly)Results against targets, what changed and why, actions
Research or analysisKey findings, implications, confidence
Evaluation or auditOverall judgement, main issues, recommendations
Project close reportOutcomes versus goals, lessons, open items
Market or competitor reportWhat changed, what it means for us, options

When to use, and when not to

Use it for any report with findings and recommendations, including recurring performance reports.

Don't use it for one-page decision memos (memos-one-pagers) or slide presentations (the Communication and Presentations skills).

Inputs

  • The question the report answers and who reads it.
  • Data, analysis, research notes and sources.
  • Targets or benchmarks for comparison.
  • Required format, length and deadline.

Raw SKILL.md file

.md

This is the exact file agents read, including the credit header and licence notice. Save it as reports/SKILL.md.

---
name: reports
description: "Write reports that get read and used: the question first, key findings with numbers, implications, evidence with headline chart titles, facts separated from interpretation, limitations and confidence stated, recommendations with owners and dates, and templates for performance, research, evaluation and project reports. Use whenever someone needs a report, monthly or quarterly performance report, research findings, analysis write-up, evaluation, audit report or end-of-project report."
license: CC-BY-4.0
metadata:
  author: Afeez Adebayo
  author_url: https://afeezadebayo.com
  author_linkedin: https://www.linkedin.com/in/afeez-adebayo/
  author_x: https://x.com/za_icon
  source: https://afeezadebayo.com/skills-md/productivity/workplace-writing/reports/
  category: productivity
  subcategory: workplace-writing
  version: "1.0.0"
  created: "2026-10-10"
  updated: "2026-10-13"
---

# Reports

Reports are written to inform decisions, yet many are read by nobody. They follow the order the work was done (method, data, then findings), bury the conclusion on page twelve, use charts without saying what they show and leave readers to work out what to do. Senior readers in particular look for the answer first and the evidence only if they doubt it.

This skill writes analytical, research, performance and project reports that lead with findings and implications, support them with clear evidence and charts that state their point, separate facts from interpretation, show limitations openly and end with recommendations and next steps. It suits monthly performance reports, research findings, evaluations, audits and end-of-project reports.

## Operating principles
1. **Answer first.** Key findings and what they mean come in the first page.
2. **Every chart has a headline.** The chart title states the point, not just the metric.
3. **Facts, interpretation and recommendation are separate.** Readers can see which is which.
4. **Show limitations.** Data gaps, sample sizes and assumptions are stated, not hidden.
5. **End with action.** Recommendations with owners and dates.

## Report types
| Type | Lead with |
|---|---|
| Performance report (monthly or quarterly) | Results against targets, what changed and why, actions |
| Research or analysis | Key findings, implications, confidence |
| Evaluation or audit | Overall judgement, main issues, recommendations |
| Project close report | Outcomes versus goals, lessons, open items |
| Market or competitor report | What changed, what it means for us, options |

## When to use, and when not to
Use it for any report with findings and recommendations, including recurring performance reports.

Don't use it for one-page decision memos (`memos-one-pagers`) or slide presentations (the Communication and Presentations skills).

## Inputs
- The question the report answers and who reads it.
- Data, analysis, research notes and sources.
- Targets or benchmarks for comparison.
- Required format, length and deadline.

## Workflow

### Step 1. Define the question
What decision or understanding the report supports, for example "Why did sign-ups fall in Q3, and what should we do?"

### Step 2. Find the answer
From the analysis, write three to five key findings, each one sentence with a number.

### Step 3. Work out implications
For each finding: so what? What does it mean for the reader?

### Step 4. Outline
Summary, findings with evidence, implications, recommendations, method and limitations, appendix.

### Step 5. Draft with evidence
Each finding gets its evidence: a chart, table or quote, with a headline that states the point.

### Step 6. Add recommendations
Specific actions with owner, date and expected effect, ranked by impact.

### Step 7. State method and limitations
Briefly: data sources, period, sample, assumptions, what the data cannot tell us.

### Step 8. Review
Check numbers against sources; test the summary with a fresh reader.

## Structure
```
1. Summary (half to one page): question, key findings, implications, recommendations
2. Findings: one section per finding with evidence
3. What it means
4. Recommendations: action | owner | date | expected effect
5. Method and limitations
6. Appendix: detailed tables, definitions, sources
```

## Chart headlines
| Weak title | Headline title |
|---|---|
| Sign-ups by month | Sign-ups fell 18% after the August pricing change |
| Support tickets Q3 | Billing questions now make up half of all tickets |
| NPS by segment | Small businesses are the least satisfied segment |

## Performance report template
```
Period: <month or quarter> | Owner: <name>
Headline: <one sentence on overall performance>
| Metric | Target | Actual | Change | Status |
What went well: <2 to 3 points>
What did not: <2 to 3 points with causes>
Actions: <owner, date>
Risks and outlook: <next period>
```

## Writing findings
| Weak | Strong |
|---|---|
| Engagement was mixed. | Weekly active users fell 9% in September, mostly among new users. |
| Costs increased. | Delivery costs rose \u00a312,000 (14%) because of the new courier contract. |
A finding states what happened, how much and, where known, why.

## Writing the summary
The summary answers four questions in half a page: What did we look at? What did we find? What does it mean? What should we do? Use numbers, avoid method detail and write it last. Many readers will read nothing else, so it must stand alone.

## Choosing charts
| Message | Chart |
|---|---|
| Change over time | Line chart |
| Comparison between groups | Bar chart, sorted |
| Part of a whole (few parts) | Stacked bar or simple table |
| Relationship between two measures | Scatter plot |
| Exact values | Table |
One message per chart; label axes and units; avoid 3D effects and unnecessary colours.

## Confidence language
| Evidence | Wording |
|---|---|
| Strong, consistent data | "The data shows..." |
| Good but limited data | "The data suggests..." |
| Early or small sample | "Early signs indicate... (sample of 40)" |
| Opinion or interpretation | "We believe... because..." |
Using consistent wording helps readers judge how much weight to give each finding.

## Recurring reports
Keep the same structure and metrics each period so readers know where to look, highlight changes since last time, and drop sections nobody uses. Ask readers once a quarter which parts help them decide and cut the rest.

## Appendices
Put detailed tables, definitions, full survey results and method notes in appendices with clear labels, so the main report stays short while specialists can still check the evidence.

## Distribution
Share the report with a short covering note: the headline finding, the decision or action needed and by when. Link to the full report rather than attaching large files, and invite questions in one place so answers reach everyone. A brief walkthrough in the next relevant meeting often helps recommendations get acted on.

## Edge cases and judgement calls
| Situation | What to do |
|---|---|
| Findings are uncertain | Give confidence levels and what would confirm them |
| Results are bad news | State them plainly with causes and a plan |
| Mixed audiences | Executive summary for leaders; detail sections for specialists |
| Very long analysis | Keep the main report short; move detail to appendices |
| Data conflicts between sources | Show both, explain the difference, choose with reasons |
| Recurring report nobody reads | Shorten to the template; ask readers what they use |

## Handoffs to related skills
- Staged writing: `writing-process`
- Summaries: `executive-summaries`
- Plain language: `plain-english`
- Feedback on drafts: `writing-feedback`
- Data analysis itself: the Data category (planned)
- Growth and marketing reports: the Growth Marketing `growth-reporting` skill

## Worked example
A quarterly customer report ran to 30 pages ordered by data source, and leaders read only the first chart. The rewrite opened with a one-page summary: three findings (churn up 2 points, concentrated in small businesses, mostly after billing problems), what they meant (about £180,000 in annual revenue at risk) and four recommendations with owners. Each finding had one chart with a headline title. Method and limitations took half a page; detailed tables moved to an appendix. Leaders discussed the recommendations in the next meeting, and two were funded.

## Common mistakes
- Method before findings.
- Charts with no stated point.
- Mixing facts and opinions.
- Hiding limitations.
- Recommendations without owners.

## Using an AI agent
The agent can draft findings from analysis outputs, write headline titles for charts, check that every number in the text matches the data, produce the summary and suggest recommendations for the user to approve. It must not invent data or overstate certainty; gaps are marked.

## Self-check before you answer
- Are key findings and implications on the first page?
- Does every chart have a headline that states the point?
- Are limitations and confidence stated?
- Do recommendations have owners and dates?

---
Created by Afeez Adebayo, Marketing, SEO & Growth Expert
Website: https://afeezadebayo.com
LinkedIn: https://www.linkedin.com/in/afeez-adebayo/
X: https://x.com/za_icon
Source: https://afeezadebayo.com/skills-md/productivity/workplace-writing/reports/

Licence: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). You may copy, share and adapt this skill, including for commercial use, as long as you credit Afeez Adebayo and link to the source URL above. Keep this notice in any copy or adapted version.

Workflow

Step 1. Define the question

What decision or understanding the report supports, for example "Why did sign-ups fall in Q3, and what should we do?"

Step 2. Find the answer

From the analysis, write three to five key findings, each one sentence with a number.

Step 3. Work out implications

For each finding: so what? What does it mean for the reader?

Step 4. Outline

Summary, findings with evidence, implications, recommendations, method and limitations, appendix.

Step 5. Draft with evidence

Each finding gets its evidence: a chart, table or quote, with a headline that states the point.

Step 6. Add recommendations

Specific actions with owner, date and expected effect, ranked by impact.

Step 7. State method and limitations

Briefly: data sources, period, sample, assumptions, what the data cannot tell us.

Step 8. Review

Check numbers against sources; test the summary with a fresh reader.

Structure

1. Summary (half to one page): question, key findings, implications, recommendations
2. Findings: one section per finding with evidence
3. What it means
4. Recommendations: action | owner | date | expected effect
5. Method and limitations
6. Appendix: detailed tables, definitions, sources

Chart headlines

Weak titleHeadline title
Sign-ups by monthSign-ups fell 18% after the August pricing change
Support tickets Q3Billing questions now make up half of all tickets
NPS by segmentSmall businesses are the least satisfied segment

Performance report template

Period: <month or quarter> | Owner: <name>
Headline: <one sentence on overall performance>
| Metric | Target | Actual | Change | Status |
What went well: <2 to 3 points>
What did not: <2 to 3 points with causes>
Actions: <owner, date>
Risks and outlook: <next period>

Writing findings

WeakStrong
Engagement was mixed.Weekly active users fell 9% in September, mostly among new users.
Costs increased.Delivery costs rose \u00a312,000 (14%) because of the new courier contract.

A finding states what happened, how much and, where known, why.

Writing the summary

The summary answers four questions in half a page: What did we look at? What did we find? What does it mean? What should we do? Use numbers, avoid method detail and write it last. Many readers will read nothing else, so it must stand alone.

Choosing charts

MessageChart
Change over timeLine chart
Comparison between groupsBar chart, sorted
Part of a whole (few parts)Stacked bar or simple table
Relationship between two measuresScatter plot
Exact valuesTable

One message per chart; label axes and units; avoid 3D effects and unnecessary colours.

Confidence language

EvidenceWording
Strong, consistent data"The data shows..."
Good but limited data"The data suggests..."
Early or small sample"Early signs indicate... (sample of 40)"
Opinion or interpretation"We believe... because..."

Using consistent wording helps readers judge how much weight to give each finding.

Recurring reports

Keep the same structure and metrics each period so readers know where to look, highlight changes since last time, and drop sections nobody uses. Ask readers once a quarter which parts help them decide and cut the rest.

Appendices

Put detailed tables, definitions, full survey results and method notes in appendices with clear labels, so the main report stays short while specialists can still check the evidence.

Distribution

Share the report with a short covering note: the headline finding, the decision or action needed and by when. Link to the full report rather than attaching large files, and invite questions in one place so answers reach everyone. A brief walkthrough in the next relevant meeting often helps recommendations get acted on.

Edge cases and judgement calls

SituationWhat to do
Findings are uncertainGive confidence levels and what would confirm them
Results are bad newsState them plainly with causes and a plan
Mixed audiencesExecutive summary for leaders; detail sections for specialists
Very long analysisKeep the main report short; move detail to appendices
Data conflicts between sourcesShow both, explain the difference, choose with reasons
Recurring report nobody readsShorten to the template; ask readers what they use
  • Staged writing: writing-process
  • Summaries: executive-summaries
  • Plain language: plain-english
  • Feedback on drafts: writing-feedback
  • Data analysis itself: the Data category (planned)
  • Growth and marketing reports: the Growth Marketing growth-reporting skill

Worked example

A quarterly customer report ran to 30 pages ordered by data source, and leaders read only the first chart. The rewrite opened with a one-page summary: three findings (churn up 2 points, concentrated in small businesses, mostly after billing problems), what they meant (about £180,000 in annual revenue at risk) and four recommendations with owners. Each finding had one chart with a headline title. Method and limitations took half a page; detailed tables moved to an appendix. Leaders discussed the recommendations in the next meeting, and two were funded.

Common mistakes

  • Method before findings.
  • Charts with no stated point.
  • Mixing facts and opinions.
  • Hiding limitations.
  • Recommendations without owners.

Using an AI agent

The agent can draft findings from analysis outputs, write headline titles for charts, check that every number in the text matches the data, produce the summary and suggest recommendations for the user to approve. It must not invent data or overstate certainty; gaps are marked.

Self-check before you answer

  • Are key findings and implications on the first page?
  • Does every chart have a headline that states the point?
  • Are limitations and confidence stated?
  • Do recommendations have owners and dates?

When to use this skill, and when not to

Use it for monthly and quarterly performance reports, research findings, analysis write-ups, evaluations, audits, market reports and end-of-project reports.

Many reports follow the order the work was done, method first and conclusion on page twelve, and senior readers never reach the answer. Charts without a stated point leave readers to guess.

Don’t use it for one-page decisions; use Memos and One-Pagers.

Example prompts

  • “Write our Q3 performance report from this data.”
  • “Turn this analysis into a report leaders will read.”
  • “Write headline titles for these charts.”
  • “Summarise our research findings with recommendations.”
  • “Shorten our 30-page monthly report.”

Example output

The skill returns a fixed structure. A shortened example for an imaginary site:

Summary: churn up 2 points, concentrated in small businesses after billing problems
Impact: about \u00a3180,000 annual revenue at risk
Findings: 3, each with a headline chart
Recommendations: 4 with owners, dates and expected effect
Method and limitations: half a page
Appendix: detailed tables

Key findings and implications sit on the first page, and every chart title states what the chart shows.

What the agent needs from you

The skill asks for the question the report answers and who reads it, the data, analysis and sources, targets or benchmarks and any required format, length or deadline.

How this skill hands off to related skills

How to run Reports step by step

  1. Install the skill as reports/SKILL.md.
  2. Define the question.
  3. Write three to five findings.
  4. Add evidence with headline charts.
  5. Add recommendations with owners.
  6. State limitations and check numbers.

Situations this skill handles

Reports handles these situations:

SituationWhat the skill does
Uncertain findingsConfidence levels stated
Bad newsPlain, with causes and a plan
Mixed audiencesSummary plus detail sections
Long analysisDetail in appendices
Conflicting dataBoth shown and explained
Unread recurring reportTemplate and reader check

What makes this skill different

Most reports mirror the analysis process. This skill reorganises around the reader's question, so findings, meaning and actions come first and evidence supports them, with limitations stated openly.

Practical tips from real projects

  • Answer first.
  • Every chart has a headline.
  • Facts, interpretation and recommendation are separate.
  • End with action.

A worked example: a quarterly customer report ran to 30 pages ordered by data source, and leaders read only the first chart. The rewrite opened with a one-page summary of three findings, churn up two points, concentrated in small businesses and mostly after billing problems, what they meant, about one hundred and eighty thousand pounds of annual revenue at risk, and four recommendations with owners. Each finding had one chart with a headline title, method and limitations took half a page and detailed tables moved to an appendix. Leaders discussed the recommendations in the next meeting, and two were funded.

Report types lead differently: performance reports with results against targets, what changed and actions; research with key findings, implications and confidence; evaluations and audits with the overall judgement, main issues and recommendations; project close reports with outcomes against goals and lessons; and market reports with what changed and what it means.

The workflow defines the question, writes three to five findings with numbers, works out the implication of each, outlines the report, drafts with evidence for every finding, adds ranked recommendations with owner, date and expected effect, states method and limitations briefly and checks every number against sources before testing the summary with a fresh reader.

Chart titles change from labels to points: sign-ups by month becomes sign-ups fell 18% after the August pricing change, and support tickets in Q3 becomes billing questions now make up half of all tickets. Chart choice follows the message: lines for change over time, sorted bars for comparisons, scatter plots for relationships and tables for exact values, with one message per chart and labelled axes.

Confidence language stays consistent: the data shows for strong evidence, the data suggests for limited data, early signs indicate with the sample size for small samples and we believe because for interpretation. Findings state what happened, how much and why, so engagement was mixed becomes weekly active users fell 9% in September, mostly among new users.

The performance report template gives the period and owner, a one-sentence headline, a metric table with target, actual, change and status, what went well, what did not with causes, actions and the outlook. Recurring reports keep the same structure each period, highlight changes and drop sections nobody uses. Reports go out with a short covering note naming the headline finding and the decision needed, link rather than attach, and gather questions in one place.

An AI agent can draft findings from analysis outputs, write chart headlines, check that every number in the text matches the data and produce the summary, but it must not invent data or overstate certainty.

Edge cases are handled openly. Uncertain findings carry confidence levels and what would confirm them, bad news is stated plainly with causes and a plan, mixed audiences get an executive summary for leaders and detail sections for specialists, very long analysis keeps the main report short with appendices, conflicting data from different sources is shown side by side with the reason for the choice made, and recurring reports nobody reads are cut back to the template after asking readers what they use.

The final check confirms that key findings and implications are on the first page, every chart has a headline, limitations and confidence are stated and recommendations have owners and dates.

A short walkthrough of the findings in the next relevant meeting helps recommendations get acted on, and questions are gathered in one place so every reader sees the answers.

Reports end with clear owners for every action.

Frequently asked questions

Summary, findings with evidence, implications, recommendations, method and limitations, then appendices.

A title that states the point, such as sign-ups fell 18% after the pricing change, rather than just the metric.

Use consistent confidence language and state sample sizes, gaps and assumptions.

As short as the decision allows, with detail moved to appendices.

A statement of what happened, how much and, where known, why.

Use a standard template, highlight changes and ask readers which parts they use.

Changelog

  • 1.0.0, 10 October 2026: first release.

Use, adapt and share this skill under CC BY 4.0: credit Afeez Adebayo and link to this page.