菜单
Use when an engineering manager needs to evaluate an engineer's level fairly, build a promotion case with evidence, or align with peers in a calibration conversation. Triggers on: "write a promotion case", "calibration prep", "level rubric design",
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/newkayak12/claude-skills --skill leveling-manager命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
正在显示 SKILL.md
基于 SOC 职业分类
| name | leveling-manager |
| effort | high |
| description | Use when an engineering manager needs to evaluate an engineer's level fairly, build a promotion case with evidence, or align with peers in a calibration conversation. Triggers on: "write a promotion case", "calibration prep", "level rubric design", |
| license | MIT |
| metadata | {"author":"wondelai","version":"1.0.0"} |
| scenarios | ["Help me write a promotion case for a Senior to Staff engineer with specific evidence","I need to prepare for calibration — how do I present my engineer's case against skepticism?","Design a level rubric for Senior and Staff engineers on my team","승진 케이스 어떻게 써야 해? 구체적인 증거 기반으로","캘리브레이션에서 내 엔지니어 승진 제안 어떻게 방어해?"] |
| compatibility | {"recommended":["think-tool"],"optional":[],"remote_mcp_note":"think-tool이 있으면 승진 케이스 작성 시 증거가 실제 해당 레벨의 기준을 충족하는지, 아니면 현재 레벨의 강한 성과인지 구분하는 판단 품질이 높아집니다. Claude 설정 → MCP Servers에서 remote SSE 엔드포인트를 추가하세요."} |
Use when:
Not for:
Promotion case structure:
## Promotion Case: [Name] — [Current Level] → [Target Level]
### Executive Summary
[1 paragraph: specific claim with named evidence. Not "she's great" — "here are three initiatives that demonstrate Staff-level behavior."]
### Evidence by Dimension
**[Dimension 1 — e.g., System Design]**
- Example 1: [STAR format]
- Example 2: [STAR format]
**[Dimension 2 — e.g., Cross-team Influence]**
- Example 1: [STAR format]
- Example 2: [STAR format]
[Repeat for each rubric dimension]
### Developmental Areas
[Honest gap statement + why it is not a blocker, or what the plan is]
| Claude | You |
|---|---|
| Generates promotion case draft from described evidence |
| Collects and validates actual evidence with direct observation |
| Writes rubric dimension definitions with observable behaviors | Makes the final call on readiness and timing |
| Prepares calibration arguments and backup examples | Advocates for the engineer in the actual calibration room |
| Drafts "not-yet" communication with specific gap + path | Has the promotion or not-yet conversation with the engineer |
| Scores your current leveling practice (0-10) | Builds the trust that makes these conversations land |
../leveling-ic/SKILL.md — understand what your ICs are working toward../1-on-1-manager/SKILL.md — 1-on-1 growth tracking is the source of continuous evidenceLeveling is not a judgment about a person — it is an assessment of demonstrated impact at a given scope. The question is never "is this person good?" It is: "At what scope does this person consistently operate, with what level of autonomy, and what evidence supports this?" Levels describe patterns of behavior over time, not single events or potential.
The foundation: Most leveling systems fail because they conflate three things: potential, performance, and level. An engineer can be performing well at their current level without being ready for the next. A high-potential engineer is not automatically at the next level until they've demonstrated it. Calibration conversations fail when managers argue about these distinctions without shared definitions. Good leveling starts with a clear rubric, gathers evidence systematically, and writes the promotion case before advocating for it.
Goal: 10/10. When evaluating your leveling practice as a manager, rate 0-10. A 10/10 means you have a documented rubric, gather evidence continuously, write promotion cases grounded in specific examples, and represent your engineers effectively in calibration. Always provide the current score and specific improvements needed.
Core concept: Every level in an engineering career ladder is defined by a combination of scope (what size problem?), autonomy (how much direction needed?), impact (what changes in the world?), and craft (how well does the technical work hold up?). A rubric translates these dimensions into specific, observable behaviors for each level.
Why it works: Without a rubric, leveling is subjective and inconsistent — different managers apply different standards, and engineers don't know what to work toward. A rubric creates a shared language: "operating at Staff level" means something specific, not "I feel like they're Staff."
Key insights:
Practice applications:
| Context | Pattern | Example |
|---|---|---|
| Level definition | Scope + autonomy + impact | Senior: "Independently owns and delivers multi-sprint projects; influences technical direction within the team; work is used as a standard by peers" |
| Rubric creation | Observable behaviors | "Staff — Technical Design: Designs systems that are adopted beyond their immediate team; identifies cross-system risks proactively; written design docs require minimal revision from principal+" |
| Sharing rubrics | Transparency | "Here's our level rubric — I want you to know exactly what we're looking for at each level" |
| Calibration alignment | Shared language | "When I say 'operates at Staff scope,' I mean the engineer regularly solves problems that span 2+ teams without being asked" |
| Level ambiguity | Name the dimension | "The disagreement is about impact, not craft. They're excellent technically — the question is whether their work influences people outside their team yet." |
Core concept: Promotion cases are won or lost in calibration based on the quality of evidence. Evidence is specific, behavioral, and tied to impact — not general impressions or tenure. Collecting evidence continuously, throughout the year, is dramatically more effective than scrambling at promotion time.
Why it works: Recency bias is the single biggest threat to fair leveling — managers remember the last three months more than the prior nine. Continuous evidence collection counters recency bias, catches growth that happened incrementally, and produces a promotion case that is hard to refute because it is grounded in facts.
Key insights:
Practice applications:
| Context | Pattern | Example |
|---|---|---|
| Running evidence doc | Weekly notes per engineer | "2026-03-12: Led the incident response for the DB outage. Coordinated across infra, backend, and SRE. RCA was clear and action items were specific. This is Staff-level cross-functional leadership." |
| STAR format | Structured evidence | "S: Q1 launch at risk due to auth service dependency. T: Block resolved without scope reduction. A: Re-architected auth integration to decouple timeline dependencies. R: Launch shipped on time; auth team adopted the pattern." |
| Soliciting input | 360 evidence | "Before I write the promotion case, I'm going to ask the tech lead of the platform team what their experience working with [name] has been" |
| Quantified impact | Numbers tell the story | "Cache layer they designed reduced infrastructure cost by 30% — $180K/year annualized. This was unsolicited and proactively identified." |
| Developmental observation | Honest gap tracking | "Design doc for the new service had scope that was too narrow — principal engineer had to expand it significantly. Not yet operating at Staff scope on system design." |
Core concept: A promotion case is a structured document that argues, with evidence, that an engineer has consistently operated at the next level. It is not a performance review, not a list of projects, and not an expression of the manager's feelings. It is an evidence-based argument, written as if a skeptic will read it.
Why it works: Calibration panels are skeptical by design — their job is to maintain level standards across the organization. A promotion case that says "she's great and ready for Senior" will fail. A case that says "here are five specific examples across six months demonstrating Staff-level behavior in three rubric dimensions" will succeed, because it's hard to refute.
Key insights:
Practice applications:
| Context | Pattern | Example |
|---|---|---|
| Executive summary | One paragraph, specific claim | "Ji-won has consistently operated at the Staff level since Q3 2025. Across three major initiatives, she has driven cross-team technical decisions, proactively identified system-level risks, and produced design patterns adopted by two other teams. The evidence below supports promotion to L6." |
| Dimension evidence | 2-3 examples per rubric area | "Technical Design (Staff criteria): (1) Auth service decoupling design — adopted by platform team. (2) Cache architecture — saved $180K/yr, identified proactively. (3) Q1 system design doc set the bar for the team." |
| Addressing gaps | Honest and bounded | "System-level influence is strongest within the backend surface; cross-stack influence (mobile, data) is still developing. This is expected at the Staff entry point and is not a blocker for promotion." |
| Skeptic test | Read it as a critic | "If I were on the calibration panel and had never seen this engineer, would this case convince me? What questions would I ask?" |
| Pre-submission review | Peer manager read | "I'm going to send this to the Staff EM on the platform team — she'll tell me if the evidence is strong enough or if I'm missing something" |
Core concept: Calibration is the process by which managers align on level standards across the organization. It is not a negotiation or an advocacy session — it is a conversation about whether a specific body of evidence meets a specific level definition. The manager's job is to represent the engineer fairly, not to win at all costs.
Why it works: Without calibration, each manager applies their own standards, and leveling becomes inconsistent. Engineers in some teams get promoted faster with less evidence; others are held to a higher bar. Calibration creates organizational fairness — the same bar applies regardless of team or manager.
Key insights:
Practice applications:
| Context | Pattern | Example |
|---|---|---|
| Evidence presentation | Lead with examples | "I'll share three specific examples that demonstrate Staff-level system design — and then I want to discuss the impact dimension" |
| Handling skepticism | Have backup examples ready | "Fair question — here's a fourth example that specifically addresses the cross-team influence criterion" |
| Yielding gracefully | Model intellectual honesty | "You're right — that example doesn't meet the Staff bar for impact. I think the case is still there, but I'll acknowledge the gap is real." |
| Challenging inconsistency | Name the standard | "We approved [engineer A] at Staff with similar evidence last cycle — I want to make sure we're applying the same bar here" |
| Post-calibration feedback | Close the loop | "Here's what calibration said: [engineer] is performing well at Senior. The specific gap for Staff is [dimension]. Here's how we'll work on that." |
Core concept: Whether the outcome is promotion, "not yet," or development feedback, the manager must communicate clearly, specifically, and with a path forward. Vague decisions ("you're not quite ready") destroy trust. Specific decisions ("the gap is system-level design, and here's what that looks like") enable growth.
Why it works: Engineers whose promotions are declined without clear rationale either leave or lose motivation. Engineers who receive specific, honest feedback about the gap — and a credible path to closing it — often grow faster than those who were promoted on schedule. The communication is as important as the decision.
Key insights:
Practice applications:
| Context | Pattern | Example |
|---|---|---|
| Promotion communication | Name what worked | "Calibration approved the promotion. What made it clear was: your leadership of the auth redesign, the cache architecture you proposed independently, and the pattern adoption across teams." |
| Not-yet communication | Specific gap + path | "Calibration said the evidence for impact at Staff scope wasn't there yet. Specifically: all strong examples are within our team. For Staff, we need to see influence outside our team. Here's how we can create that opportunity in Q2." |
| Gap communication | Concrete and observable | "The gap isn't technical skill — it's the scope of problems you're choosing. Staff engineers identify and propose solutions to problems before anyone asks them to. Let's work on that." |
| Timeline setting | Realistic and committed | "I think you could be ready for the next cycle — that's about 6 months. Here are the three things I want to see between now and then." |
| Written follow-up | Document the conversation | Add to 1-on-1 doc: "Promotion decision (2026-03): Calibration approved. Key evidence: [list]. Growth focus for Staff+: [dimension]." |
| Mistake | Why It Fails | Fix |
|---|---|---|
| Promoting on potential, not demonstrated behavior | Sets the engineer up to fail; creates a credibility problem for the manager | Require evidence of consistently operating at the next level, not just glimpses |
| Promotion case built on recency | Last 3 months dominate; 9 months of evidence is lost | Keep running evidence notes year-round; reference them at promotion time |
| Vague promotion cases | Calibration panel can't evaluate; skeptics win | Every dimension requires 2-3 specific examples; no general impressions |
| Advocating without listening in calibration | Misses real gaps; damages credibility with peers | Listen to objections; have additional evidence ready; yield when the gap is real |
| Not sharing rubrics with engineers | Engineers can't develop toward unknown criteria | Share rubrics in 1-on-1s; review them together; let engineers self-assess |
| Sugarcoating a not-yet decision | Engineer doesn't change behavior; repeat cycles of disappointment | Be specific and honest about the gap; provide a concrete path forward |
| Conflating performance and level | High performer at L4 gets L5 promotion before demonstrating L5 scope | Keep performance and leveling discussions separate; both matter, neither substitutes for the other |