HR pack
Claude Skill
Performance Review Drafter
Drafts honest, specific perf reviews from notes — avoids vague platitudes.
What it does
Given notes on the employee's work over the review period (wins, struggles, peer feedback, project outcomes), produces a perf review draft with specific examples, concrete growth areas, and a clear overall narrative. Resists "diplomatic" language that hides real signal.
When to use
- ✓You have notes/data and need to turn them into a coherent review
- ✓You're tempted to write something vague to avoid hard conversations
- ✓Multiple peers gave feedback and you need to synthesize signal vs noise
When not to use
- ✗You don't have actual examples — vague reviews are worse than no reviews
- ✗You haven't had any 1:1 conversations about the issues — don't surprise people in writing
Install
Download the .zip, then unzip into your Claude skills folder.
mkdir -p ~/.claude/skills
unzip ~/Downloads/performance-review-drafter.zip -d ~/.claude/skills/
# Restart Claude Code session.
# Skill is now available — Claude will use it when relevant.SKILL.md
SKILL.md
---
name: performance-review-drafter
description: Use when drafting a performance review based on notes, peer feedback, or project outcomes. Triggers on "draft a perf review", "performance review for", "year-end review", "mid-year review".
---
# Performance Review Drafter
Most performance reviews are useless because they're vague. "Strong performer, continue to grow" tells the employee nothing. Build reviews that someone reading them in 6 months can act on.
## Required inputs
1. **Employee role and level**
2. **Review period** (mid-year / annual / quarterly)
3. **Specific wins** — projects shipped, outcomes, customer wins
4. **Specific struggles** — what didn't go well, with context
5. **Peer feedback** (if collected) — quoted or paraphrased
6. **Manager's overall verdict**: exceeds / meets / below
If the manager only has vibes ("they were fine"), push back: "What's a specific moment in the last 6 months where they did great work? Where they fell short?"
## Structure
### Overall narrative (3-4 sentences)
Open with the verdict. Be direct.
- "Strong performer this period. Specifically excelled at X, with growth needed in Y."
- Avoid: "[Name] continued to grow and develop in their role."
### Strengths (3-4 bullets, each with evidence)
- **Pattern, then example.** "Strong technical judgment — when the API gateway started timing out in March, [name] quickly identified the root cause as connection pool exhaustion and fixed it before it impacted customers."
- One bullet per strength, NOT a laundry list
### Growth areas (2-3 bullets, each actionable)
- Frame as behaviors, not personality. "Communicates technical decisions more proactively in cross-functional reviews" beats "needs to communicate better."
- Each one needs a "what good looks like" anchor.
- Bad: "Could be more proactive."
- Good: "Could surface blockers earlier in the project — for the X migration, the team didn't know about the schema dependency until 2 weeks before launch."
### Goals for next period
- 2-3 concrete goals tied to growth areas
- Each measurable or observable
### Optional: peer feedback synthesis
If peer feedback is available, synthesize themes (don't quote individuals). Note where peer feedback DIVERGES from manager view — that's high-signal.
## Tone calibration
For **strong performer**:
- Lead with the strongest signal of impact
- Growth areas should still be specific (no fake humility)
- "What's the next role this person could grow into?"
For **solid / meets**:
- Don't pad with strengths to soften
- Growth areas should be the substance
- Be honest about what would move them to "exceeds"
For **underperforming**:
- Direct, with examples
- Don't soften with hedge words ("sometimes", "occasionally")
- Make clear what would change the trajectory
- This review should not be a surprise — they should have heard the substance in 1:1s already
## Anti-patterns to strip
- "Continues to grow" — meaningless filler
- "Strong contributor" — vague
- "Could be more X" — passive; rephrase as observable behavior
- Diplomatic doublespeak that obscures the actual verdict
- Peer feedback used as cover ("some peers mentioned…") — own the feedback or leave it out
## Output format
Markdown. Quote specific examples. Use peer-feedback themes (not direct quotes that could be identified).
Example prompts
Once installed, try these prompts in Claude:
- Draft a mid-year review for [name], Senior Engineer. Wins: shipped X migration on time, mentored junior. Struggles: code reviews slow, conflict avoidance with PM. Peer feedback: technical depth strong, communication unclear in incidents.
- Draft a year-end review for an underperforming PM. Need to be honest but constructive — they're not on a PIP yet but trending.