A Claude Code skill that generates well-structured OKRs from strategic context, or reviews and rewrites existing OKRs to fix the common failures that make them useless.
Two modes in one skill:
Generate — Takes your team's context (product area, strategic priorities, timeframe, team size) and drafts 1–3 objectives with 3–4 key results each. Each KR includes a baseline → target format and a calibration note on whether the set is realistic for your team's capacity.
Review — Takes your existing OKRs and scores each objective and KR against clear criteria. Names the anti-pattern when it finds one, rewrites weak KRs on the spot, and gives an overall verdict on the set.
Product managers writing quarterly or half-yearly OKRs, or teams that keep ending up with task lists dressed up as key results. Also useful before OKR review season when you want to pressure-test a draft before leadership sees it.
Generate mode:
- 1–3 objectives with 3–4 KRs each
- Baseline → target format for each KR
- Calibration note on capacity and ambition
Review mode:
- Score and verdict per objective and KR
- Named anti-pattern for each weak KR
- Rewrite of every KR that doesn't pass
- Overall verdict on the set
Install via Claude Code. Clone this repo and copy the folder into your skills directory:
macOS / Linux:
cp -r ProductManagerOKRWriter ~/.claude/skills/okr-writerWindows:
Copy-Item -Recurse ProductManagerOKRWriter "$env:USERPROFILE\.claude\skills\okr-writer"Restart Claude Code. Then trigger with natural language:
To generate:
- "Help me write OKRs for my team — we own activation and onboarding, Q3"
- "Draft OKRs for our AI features team this half"
To review:
- "Review my OKRs — I think some of them are tasks not results"
- "Are these OKRs good?" + paste your OKRs
- "My KRs feel like tasks, can you fix them?"
| Anti-pattern | Example |
|---|---|
| Task KR | "Launch the new onboarding flow" |
| Vanity metric | "Increase page views by 20%" |
| Output as outcome | "Ship 3 new features" |
| Too many KRs | 6+ KRs per objective |
| Unmeasurable objective | "Make the product more delightful" |
| Unanchored target | "Achieve 90% NPS" with no current NPS stated |
Three test cases in evals/:
- Generate OKRs for an activation/onboarding team — straightforward strategic context, tests baseline → target format
- Review weak OKRs with common anti-patterns — task KRs, vanity metrics, unmeasurable objectives
- Generate OKRs for an AI product with measurement challenges — tests how the skill handles subjective quality metrics