amazon-memo
Skill172 starsChanged 4 months agoMITView on GitHub
Create, adapt, or revise Amazon-style narrative decision memos, six-pagers, PR/FAQs, and short decision briefs, including reading meetings and async review. Use for choosing the right memo form, evidence-led reasoning, and revising mechanical AI prose for human readers. Do not use for general copywriting, document rendering, or replacing incident reports and technical design records.
In this repository192 other files it gives its agents.
- agent-production-operations/AGENTS.mdAGENTS.md
- AGENTS.mdAGENTS.md
- forward-deployed-engineering/AGENTS.mdAGENTS.md
- production-excellence/AGENTS.mdAGENTS.md
- product-lifecycle/AGENTS.mdAGENTS.md
- tailscale/AGENTS.mdAGENTS.md
- workflow-architect/AGENTS.mdAGENTS.md
- llms.txtllms.txt
- actuarial-risk-modelingSkill
- adr-authoringSkill
- agent-councilSkill
- agent-evals-and-observabilitySkill
- agent-production-operationsSkill
- agent-skillsSkill
- ai-governanceSkill
- ai-operating-economicsSkill
- analog-occultismSkill
- anydocSkill
- api-design-and-evolutionSkill
- artifact-pyramidsSkill
- ascii-city-engineSkill
- autogenSkill
- backend-engineeringSkill
- binary-analysisSkill
- bmadSkill
- brand-designerSkill
- c4-diagrammingSkill
- camera-hackingSkill
- capacity-and-cost-engineeringSkill
- chief-of-staff-methodologySkill
- cli-builderSkill
- cncf-landscapeSkill
- code-reviewSkill
- color-managementSkill
- community-guidesSkill
- conditional-customer-successSkill
- confluence-cliSkill
- constrained-optimizationSkill
- crewaiSkill
- crmSkill
- crowdsecSkill
- cryptpadSkill
- cyberpunkSkill
- daily-life-discoverySkill
- data-architectSkill
- data-cleaningSkill
- data-engineeringSkill
- data-scientistSkill
- de-spinSkill
- digital-twinSkill
- docker-composeSkill
- documentsSkill
- dsm5Skill
- dspySkill
- electronicsSkill
- emailSkill
- enterprise-architectureSkill
- epubSkill
- esp32-developmentSkill
- ffmpegSkill
- The first 60 shown.
The file
We split this skill into the rules it gives an agent
12 rules. About 0 are common, 9 are distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈9
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 21
Use adr-authoring for durable architecture decisions and incident-learning for incident analysis.
- Very specific: line 24
Use technical-documentation for instructions and runbooks, writers-helper for general manuscript craft, and documents for rendered office …
- Very specific: line 26
Use product-discovery or product-methodology when the missing work is understanding the problem or testing demand.
- Distinctive: line 38
Use the routing table below.
- Distinctive: line 54
Keep decision-critical caveats in the body.
- Distinctive: line 56
Revise for human reading: Read the revision tests for every draft or revision.
- Distinctive: line 85
If actual compliance matters, declare paper size, font, margins, spacing and appendix treatment, then render and inspect.
- Distinctive: line 88
Recommendations are not corporate rules: Read sources and claim boundaries when explaining provenance or making claims about a company's …
- Distinctive: line 91
Persuasiveness must preserve truth: If asked to make a weak case sound certain, strengthen the reasoning and state the gaps.
- Distinctive: line 97
Read synthetic worked examples when the user needs an illustration or the draft suffers from inflated claims or abstract prose.
- Distinctive: line 99
Every example is invented and generic.
- Distinctive: line 109
Stop when the artifact supports its stated reading and decision goal.
≈ “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
Amazon Memo
Help a reader understand a problem, challenge the reasoning, and make or record a decision. Narrative means connected reasoning in prose, not decorative storytelling. Amazon's conventions are a useful starting point; respect the receiving team's own written culture. These are original scaffolds, not official Amazon templates.
When not to use
- Use
adr-authoringfor durable architecture decisions andincident-learningfor incident analysis. Add an executive cover memo only when useful; retain the underlying artifact's structure and evidence. - Use
technical-documentationfor instructions and runbooks,writers-helperfor general manuscript craft, anddocumentsfor rendered office files. - Use
product-discoveryorproduct-methodologywhen the missing work is understanding the problem or testing demand. A polished memo cannot supply that evidence.
Workflow
- Establish the reading contract. Identify the reader, decision or learning goal, decision owner, deadline, known local conventions, and meeting length or async window. Use supplied context before asking. If a missing detail changes the recommendation, ask one focused question; otherwise state a reasonable assumption and draft. Do not turn intake into an interview by default.
- Choose and explain the form in one sentence. Respect an explicitly requested artifact; explain a mismatch and offer a better fit without silently substituting. Use the routing table below. Read form and length guidance when choosing between forms, planning pages, or adapting to another organization.
- Build the argument before polishing. State the customer or stakeholder problem, relevant baseline, credible alternatives (including continuing as-is), proposed mechanism, costs, risks, uncertainty, and what decision is needed. Separate observed facts, assumptions, forecasts, and invented illustration. Keep citations near claims; identify missing denominators, dates, costs, or sources instead of inventing them. Preserve contradictory findings and explain what they change. With sparse evidence, propose a bounded test or conditional decision rather than a confident rollout. For a test, specify proposed scope, duration or review date, owner, and outcome evidence to inspect. Label unsupplied dates and thresholds as proposals; do not leave the checkpoint as an indefinite request for more evidence.
- Draft for this audience. Load only the selected template. Use paragraphs to connect evidence to consequences. Use headings, compact tables, lists, charts, and appendices when they reduce reader effort; do not turn slides into bullet walls. Define specialist terms for unfamiliar readers and retain precise terms for experts. Keep decision-critical caveats in the body. Templates are starting points; omit irrelevant headings and add what the decision actually needs.
- Revise for human reading. Read the revision tests for every draft or revision. Run the argument, evidence, paragraph, and read-aloud passes. Preserve the writer's voice and legitimate technical language. Revise actual reasoning and sentences; do not use banned-word lists, fixed sentence lengths, made-up readability scores, or synonym swapping as a substitute for judgment.
- Design review and closure. For a meeting or async review, read reading and decision practices. Name the decision owner, reading arrangement, questions to resolve, and where the outcome will live. Allow dissent; do not manufacture consensus. Writing a memo does not authorize sending it, publishing it, or scheduling a meeting.
Form routing
| Main reader need | Default form and reason | Load when drafting |
| Choose a strategy, resource allocation, staffing plan, or operating change | Decision/strategy narrative: compare choices and consequences | Narrative template |
| Explore a future customer experience, product, service, or process | PR/FAQ: work backwards from value to hard feasibility questions | PR/FAQ template |
| Decide a narrow, bounded question with limited context | Short decision brief: keep decision overhead proportionate | Brief template |
| Inspect operating results, learn from an incident, approve technical design, or execute a procedure | Retain the business-review, incident, design, ADR, or procedure form; add narrative rationale where useful | Form boundaries |
Non-negotiable distinctions
- Six pages is a ceiling, not a requirement. Amazon's 2024 shareholder letter describes a maximum six-page narrative; do not pad two pages of substance to six. Local rules may differ. No universal font, word count, or heading sequence is established here.
- A press release is not the whole PR/FAQ. The future release states customer value; customer and internal FAQs test the proposal. Mark all imagined quotes and future claims as hypothetical. Never present them as testimony or shipped capability.
- Page estimates are not pagination. Markdown cannot prove a rendered page count. If actual compliance matters, declare paper size, font, margins, spacing and appendix treatment, then render and inspect. Never shrink type or move essential risks out of the body to make the count fit. Label unrendered lengths as estimates.
- Recommendations are not corporate rules. Read sources and claim boundaries when explaining provenance or making claims about a company's practices. Attribute practitioner guidance. Independent written cultures are not presumed Amazon adoptions.
- Persuasiveness must preserve truth. If asked to make a weak case sound certain, strengthen the reasoning and state the gaps. Do not inflate demand, suppress counterevidence, invent numbers, or conceal uncertainty.
Supporting material
Read synthetic worked examples when the user needs an illustration or the draft suffers from inflated claims or abstract prose. Every example is invented and generic. Do not reuse personal, client, interview, or confidential material in public examples.
Completion and exit conditions
Deliver the requested draft or revision, a brief form rationale when relevant, material assumptions and evidence gaps, and the review/decision arrangement if requested. Report actual pagination status honestly. For a revision, briefly explain the substantive changes rather than claiming an objective style score.
Stop when the artifact supports its stated reading and decision goal. If evidence cannot support the requested conclusion, deliver a qualified draft or an evidence request with the consequence for the decision. After two non-converging revision passes, name the unresolved audience or argument choice instead of endlessly polishing.
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.

