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",
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
A direct command skips the review prompt. Inspect the source before running it.
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."}
Standing Mandates
ALWAYS distinguish level evidence from performance evidence — strong performance is not level evidence.
ALWAYS prepare counter-arguments to anticipated calibration objections before the meeting.
NEVER write a promotion case with general claims — every assertion needs a specific example.
NEVER evaluate level by comparing to a single highlight; evaluate against the full rubric.
Engineering Leveling — Manager Framework
When to Use / When Not to Use
Use when:
Writing or reviewing a promotion case before a calibration cycle
Designing or sharing level rubrics with your engineers
Preparing to advocate for a promotion in a calibration session
Performance improvement plans (different from leveling)
Process
Define rubric — ensure rubric exists and is shared with the engineer
Collect evidence continuously — running doc per engineer; STAR format; tag by rubric dimension
Write the promotion case — structure: executive summary + evidence by dimension (2-3 examples each) + developmental areas
Apply the skeptic test — read the case as someone who doesn't know this engineer; find the weak spots
Calibration — present evidence, not opinions; have additional examples ready; know when to yield vs. push back
Communicate — promotions: name what demonstrated readiness; not-yet: specific gap + path + timeline
Output Template
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]
What Claude Does / What You Do
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
Related Skills
../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 evidence
Core Principle
Leveling 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.
Scoring
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.
9-10: Clear rubric shared with engineers, evidence gathered continuously, promotion cases written with specific examples, calibration participation is strong
7-8: Rubric exists but isn't shared proactively; promotion cases are written but rely on recency; calibration is reactive
5-6: Leveling is intuition-based; no rubric; promotion cases are thin; calibration is uncomfortable
3-4: Leveling is opaque to engineers; promotions happen based on tenure or politics; no calibration
1-2: No leveling framework; "good" vs "not good" is the only assessment; engineers don't know what's expected
The Leveling Manager Framework
1. Level Definitions and Rubrics
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."
The key inflection points: Junior→Mid (independence within tasks), Mid→Senior (independence within projects), Senior→Staff (influence across teams/systems)
Rubric dimensions: Technical Execution, System Design, Project Leadership, Collaboration, Impact
Each dimension should have 3-4 observable behaviors per level — specific enough to find evidence for
Share the rubric with engineers: if they don't know what "Senior" looks like, they can't develop toward it
Rubrics should be calibrated across managers, not defined by one person
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."
2. Evidence Collection
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:
Keep a running doc per engineer: weekly or bi-weekly entries of specific examples with context
After every significant event (incident, design review, launch, conflict), write a note: "What did they do? What was the impact? What does this demonstrate about their level?"
Use the STAR format: Situation → Task → Action → Result
Collect both positive evidence (behaviors at the target level) and developmental observations (gaps at the target level)
Ask other managers, engineers, and stakeholders for input — don't be the only observer
Quantify impact where possible: "The refactor they led reduced API latency by 40%, enabling the SLA upgrade for three enterprise customers"
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."
3. Writing the Promotion Case
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:
Structure: Executive summary (1 paragraph) → Evidence by rubric dimension (3-5 paragraphs) → Summary of level rationale → Developmental areas
Each rubric dimension needs at least 2-3 specific examples — no dimension should be supported by a single data point
Write the executive summary last — it should synthesize the evidence, not preview vague impressions
Address the developmental areas honestly: "Not yet at full scope on [dimension] — here's the plan" is more credible than pretending there are no gaps
Write it as a document, not a list of bullet points — narrative is more persuasive and easier to follow
Have another senior manager review it before submission — they'll see the gaps a sponsor can't
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"
4. Calibration Conversations
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:
Know the level rubric cold before calibration — you need to speak to evidence against specific criteria
Present evidence, not opinions: "Here are three examples of Staff-level behavior" not "I really believe she's ready"
Be prepared for skepticism — have additional examples ready for the dimensions that will be questioned
Know when to yield: if calibration reveals a genuine gap you missed, update your assessment — that's the system working
Know when to push back: if calibration is applying an inconsistent standard, name it: "What would constitute sufficient evidence? Can we align on the bar?"
Document calibration outcomes and share with engineers (appropriately) — they should know what was discussed and what would change the outcome
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."
5. Communicating Leveling Decisions
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:
Promotions: name what demonstrated readiness — "what made this promotion clear was X, Y, Z"
Not-yet decisions: name the specific gap, not a vague feeling — "the gap is X, which at Staff level looks like Y"
Provide a timeframe and a plan: "Here's what I want to see in the next two quarters" + specific opportunities
Don't sugarcoat a gap you privately believe is uncloseable — if the engineer has a genuine ceiling, have that harder conversation
Separate the decision from the relationship: "I advocated for you; the calibration panel identified a gap that I think is worth discussing"
Follow up in writing: decisions and plans should be in the shared 1-on-1 doc, not just spoken
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]."
Common Mistakes
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
