eol-stakeholder-sequence
정보
이 Claude Skill은 제품 서비스 종료(EOL)를 위한 이해관계자 대화의 순서와 내용을 계획합니다. 법무팀, 재무팀, 매출 관련 팀, 고객 대면 팀 순서로 반드시 진행해야 하는 중요한 순서를 강제하여 공식 발표 전에 위험 요소를 발견합니다. EOL 결정이 내려졌을 때 사용하면, 체계적으로 잠재적 이의 사항을 찾아내고 각 대화에서 다뤄야 할 내용을 정의할 수 있습니다.
빠른 설치
Claude Code
추천npx skills add deanpeters/Product-Manager-Skills -a claude-code/plugin add https://github.com/deanpeters/Product-Manager-Skillsgit clone https://github.com/deanpeters/Product-Manager-Skills.git ~/.claude/skills/eol-stakeholder-sequenceClaude Code에서 이 명령을 복사하여 붙여넣어 스킬을 설치하세요
문서
EOL Stakeholder Sequence
Purpose
Plan the order of EOL conversations, and what each one must cover. The output is a sequenced list of stops — each with what you need from them, what you owe to them, the red flags to listen for, and what the conversation must produce.
EOL is not a broadcast. It is a series of conversations where each one informs the next, and each one surfaces something the last one missed. Sequenced well, the awkward discoveries happen in a conference room. Sequenced badly, they happen in public, after the announcement, in front of customers.
Input
Works best with: The product being sunset and who already knows about it.
Also useful: Scale (customers, revenue, contracts), whether channel partners or regulators are involved, and any political sensitivities, strained relationships, or past surprises worth planning around.
Anything supplied with the invocation itself — text after the skill name, a pasted context dump, or
an appended ARGUMENTS: line — counts as answers already given. Use it and skip whatever it
covers; don't re-ask.
Arriving empty-handed? That works too. The skill asks up to three questions — what's being sunset and who knows, how big the blast radius is, and whether there's history to plan around — then builds the sequence. Political sensitivities are optional; the sequence works without them and sharpens with them.
Example invocations:
Sequence the stakeholder conversations for retiring our legacy reporting module.Who do I talk to, in what order, about sunsetting a hardware line with 8 resellers?
Key Concepts
The Sequencing Principle
Talk to the people who can kill the plan before you talk to the people who have to execute it.
Each conversation should inform the next. Legal exposure first, because a contract term can end the discussion. Financial impact second, because it sizes everything downstream. Then revenue-facing teams, then customer-facing teams, then the technical teams who will carry it out.
Get this backwards — brief Support first, Legal last — and you will have told forty people about a plan that a single contract clause invalidates.
Right-Size the Sequence
Not all EOLs play out the same. Most land in the middle:
| Level 1 — Light | Level 2 — Standard | Level 3 — Heavy | |
|---|---|---|---|
| Typical scope | Feature, internal tool, API | Commercial product, active customers | Revenue-critical, hardware, regulated |
| Stops | 3-4 | 7-8 | 10+ |
| Includes | Engineering, Support, affected users | + Legal, Finance, Sales, Marketing, CS, difficult customers | + Executives, Channel, Regulatory, key accounts |
| Skip unless there's a reason | Legal, Finance, Sales, Channel | Channel, Regulatory | — |
Level 2 is the default. Recommend a level, say why, then let the user move it. Never default to the heaviest sequence — a ten-stop tour for a feature deprecation burns credibility you'll want for the sunset that actually needs it.
If someone dials down, name the stop they're dropping and what it typically catches. Dropping Legal on a product with contracts is the one worth pushing back on once — then honoring their call.
The Canonical Order
Filtered by level, but the relative order holds:
Level 1+
- Engineering — what depends on this technically
- Support — what changes for support operations
- Affected users or internal teams — who feels it first
Level 2+ (inserted ahead of the above)
- Legal — contractual and regulatory exposure
- Finance — revenue impact and forecast changes
- Sales — pipeline, bundles, and promises made in the field that never found their way into a contract or a ticket
- Marketing — you do not want to discover they just bought a quarter's worth of demand gen for the thing you killed
- Customer Success — they will bear the brunt of this; work with them, not around them
- Your most difficult customers — they will find the three things you forgot that would have blown up six months later
Level 3+
- Executive leadership — strategic approval
- Channel partners — inventory, commitments, and their own customer relationships
- Regulatory bodies — filings and compliance
- Key accounts — individual transition conversations
Why Your Most Difficult Customers Belong in the Sequence
Counterintuitive, and the stop teams skip most often. The customer who files the most tickets and pushes hardest in QBRs has, by construction, the deepest and weirdest usage of your product. They will find the integration you forgot, the contract term nobody read, and the workflow that has no equivalent in the replacement.
You can learn that from them in a scheduled call now, or from them on a public forum later.
What You Need / What You Owe
Every stop has two directions. Skipping the second turns stakeholders into an audience:
- What you need FROM them — the input, approval, or knowledge only they have
- What you owe TO them — the context, commitment, or lead time they need to do their job
A conversation that is all extraction gets you compliance. A conversation that runs both ways gets you the thing they weren't asked about.
The Sticky-Note Rule
Every bullet is 4 to 8 words. Sequences get read in hallways, before meetings, on phones.
Anti-Patterns (what this is NOT)
- Not a stakeholder map. No power/interest grid. This is an ordered itinerary.
- Not a broadcast plan. The customer announcement is a different artifact.
- Not parallelizable by default. The order is the value; parallelize only where noted.
- Not a RACI. Decision rights are a separate question from conversation order.
Application
Step 1: Set the level
Recommend from blast radius, present all three, let the user choose. If unclear, ask:
"How complex is this sunset?
- Light — a feature or internal tool. Few users, all reachable.
- Standard — a commercial product with paying customers and contracts.
- Heavy — revenue-critical, hardware, channel partners, or regulated.
Most are Standard. Which sounds like yours?"
Step 2: Order the stops
Apply the canonical order, filtered by level. Two adjustments worth making:
- Move a stop earlier when history demands it. If Sales was blindsided last time, brief Sales earlier than the template says and note why.
- Split a stop when one function holds two different conversations. "Legal (contracts)" and "Legal (regulatory)" are often different people with different answers.
Step 3: Build each stop
For every stop, fill five fields:
- When — before or after which milestone or other stop
- Why this stop matters for EOL — in 4-8 words
- What you need FROM them — specific questions, not "input"
- What you owe TO them — context, lead time, commitments
- Red flags to watch for — the signals that this stop just surfaced a blocker
Then state the output: the decision, approval, or artifact the conversation must produce. A stop with no output is a status update, not a stop.
Step 4: Mark parallel vs. sequential
Say which stops must be strictly sequential and why, and which can safely run in parallel. Being explicit here is what lets a team compress a ten-stop sequence into three weeks without breaking the dependency that mattered.
Step 5: Assumptions
Close with what you assumed — who you think knows, what you think the contracts say, which relationships you think are intact.
Final Step: Offer what comes next
"Where next?
- Build the phase-gated checklist — the operational plan (Recommended)
- Draft talking points for the three most sensitive stops
- Build the internal enablement pack — support FAQ, sales talking points, objections
- Draft the customer announcement — see
eol-message
Reply with a number, a combination ('1 & 3'), or your own path."
Examples
examples/sample.md— Fieldlight Classic Dispatch (SaaS, Level 2, eight stops)examples/sample-industrial.md— NFA-200 controller line (industrial, Level 3, eleven stops)
Common Pitfalls
Pitfall 1: Announcing Before Asking
Symptom: The sequence starts with a company-wide email.
Consequence: Every subsequent conversation is a negotiation with someone who found out the wrong way and is now defending their turf instead of helping you.
Fix: The announcement is the end of the internal sequence, not the beginning.
Pitfall 2: Legal Last
Symptom: Legal gets looped in once the plan is built, "to review."
Consequence: A contract term or regulatory obligation invalidates a plan forty people have already committed to. Now you're relitigating instead of executing.
Fix: Legal is the first Level 2+ stop. A five-minute "are there commitments here?" beats a five-week replan.
Pitfall 3: Treating Sales as Recipients
Symptom: Sales is informed of the EOL date rather than asked what they've promised.
Consequence: Field commitments that never made it into a contract surface after the announcement, from the customer's side, as accusations of bad faith.
Fix: Ask Sales explicitly: what has anyone promised about this product's future? Ask before you tell.
Pitfall 4: Skipping the Difficult Customers
Symptom: The sequence covers internal functions and stops.
Consequence: Your edge cases are discovered by the people least inclined to be quiet about them.
Fix: Two or three of your hardest customers, under NDA if needed, before the public announcement. Budget for what they'll find.
Pitfall 5: Extraction-Only Conversations
Symptom: Every stop is a list of things you need. Nothing is offered.
Consequence: Stakeholders answer the questions asked and volunteer nothing. The thing that sinks you was the thing nobody thought to ask about.
Fix: Fill "what you owe them" for every stop before you hold it.
References
Related Skills
These stand on their own — none is a prerequisite for this skill, and this skill isn't a prerequisite for them. If you already chose a level elsewhere, say "Level 2" and this skill builds to it.
eol-checklist— the phase-gated operational planeol-internal-enablement— what customer-facing teams get after these conversationseol-readiness-advisor— if the decision still needs a caseeol-message— the customer-facing announcementstakeholder-map— general power/interest mapping for non-EOL work
External Frameworks
- Mendelow's Matrix — power/interest prioritization, useful for choosing depth per stop
- DACI / RACI — decision rights, a separate question from conversation order
Provenance
- Adapted from
prompts/eol-stakeholder-sequence.mdin thehttps://github.com/deanpeters/product-manager-promptsrepo.
GitHub 저장소
자주 묻는 질문
eol-stakeholder-sequence Skill이란 무엇인가요?
eol-stakeholder-sequence은(는) deanpeters이(가) 만든 Claude Skill입니다. Skill은 Claude가 필요할 때 불러오는 지침과 리소스를 묶어 추가 프롬프트 없이 eol-stakeholder-sequence 관련 작업을 수행할 수 있게 합니다.
eol-stakeholder-sequence은(는) 어떻게 설치하나요?
이 페이지의 설치 명령을 사용하세요. eol-stakeholder-sequence을(를) Claude Code 플러그인으로 추가하거나 저장소를 skills 디렉터리에 복제한 다음 Claude를 다시 시작해 Skill을 불러옵니다.
eol-stakeholder-sequence은(는) 어떤 카테고리에 속하나요?
eol-stakeholder-sequence은(는) 기타 카테고리에 속합니다.
eol-stakeholder-sequence은(는) 무료로 사용할 수 있나요?
네. eol-stakeholder-sequence은(는) AIMCP에 등록되어 있으며 무료로 설치할 수 있습니다.
연관 스킬
LlamaGuard는 폭력 및 혐오 발언 등 6가지 안전 범주에서 LLM 입력과 출력을 조정하기 위한 Meta의 70-80억 파라미터 모델입니다. 94-95% 정확도를 제공하며 vLLM, Hugging Face 또는 Amazon SageMaker를 사용해 배포할 수 있습니다. 이 기술을 사용하여 AI 애플리케이션에 콘텐츠 필터링 및 안전 가드레일을 손쉽게 통합하세요.
이 Claude Skill은 리소스 적정화, 태깅 전략, 지출 분석을 통해 개발자들이 클라우드 비용을 최적화할 수 있도록 지원합니다. AWS, Azure, GCP에서 클라우드 비용을 절감하고 비용 거버넌스를 구현하기 위한 프레임워크를 제공합니다. 인프라 비용을 분석하거나, 리소스를 적정화하거나, 예산 제약을 충족해야 할 때 사용하세요.
이 Claude Skill은 스프레드, 오버/언더, 프로프 베트를 포함한 스포츠 베팅 시장을 분석합니다. 역사적 추이와 상황별 통계를 검토하여 가치 베트를 발견하고, 교육적 목적으로 실행 가능한 권장 사항이 담긴 구조화된 마크다운 결과를 제공합니다. 개발자는 이 기능을 스포츠 베팅 분석 도구에 활용할 수 있으며, 단순히 엔터테인먼트/교육 목적으로만 설계되었음을 유의해야 합니다.
이 스킬은 bitsandbytes를 사용하여 LLM을 8비트 또는 4비트 정밀도로 양자화하며, 최소한의 정확도 손실로 50-75%의 메모리 감소를 달성합니다. 제한된 GPU 메모리에서 더 큰 모델을 실행하거나 추론을 가속화하는 데 이상적이며, INT8, NF4, FP4와 같은 형식을 지원합니다. 이 스킬은 HuggingFace Transformers와 통합되어 QLoRA 학습 및 8비트 옵티마이저를 가능하게 합니다.
