reference-class-forecasting — Claude Code skill: how it works
reference-class-forecasting grounds predictions in historical track records by finding similar past events and establishing baseline probabilities before judging specific cases.
no license 105 installs ★ 203 updated 2026-09-01
npx skills add lyndonkl/claude --skill reference-class-forecasting -a claude-codeWhat reference-class-forecasting does
The skill structures forecasts around the outside view methodology. Instead of relying on gut feelings, it prompts the assistant to establish a reference class of prior outcomes and derive a baseline frequency. When published numbers are absent, it uses starter anchors such as 10–20% for novel concepts or 70–80% for established processes.
It guides users through multi-stage funnels by multiplying probabilities across sequential gates. It also stress-tests claims of special circumstances, calculating how much evidence is needed when someone departs by more than 10% or 30% from the base rate. Historical sets with fewer than 20 cases trigger a warning to broaden the scope.
What it helps with
- determine an empirical baseline probability for a project
- audit claims that a specific case breaks historical patterns
- calculate combined odds across sequential multi-step funnels
- verify whether a comparison dataset has sufficient historical sample size
Example prompts for the agent
- Find the historical base rate for launching a B2B SaaS startup
- Stress-test my assumption that our project will beat past delivery timelines
- Calculate the sequential funnel probability for passing this legislative bill
How to install reference-class-forecasting
1Skills CLI (Vercel)
npx skills add lyndonkl/claude --skill reference-class-forecasting -a claude-code2Manually: put the files in place
~/.claude/skills/reference-class-forecasting/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/reference-class-forecasting/
1Skills CLI (Vercel)
npx skills add lyndonkl/claude --skill reference-class-forecasting -a codex2Manually: put the files in place
.agents/skills/reference-class-forecasting/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add lyndonkl/claude --skill reference-class-forecasting -a cursor2Manually: put the files in place
.cursor/skills/reference-class-forecasting/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install lyndonkl/claude skills/reference-class-forecasting2Manually: put the files in place
.github/skills/reference-class-forecasting/SKILL.md1Skills CLI (Vercel)
npx skills add lyndonkl/claude --skill reference-class-forecasting -a gemini-cli2Platform CLI
gemini skills install https://github.com/lyndonkl/claude --path skills/reference-class-forecastingManually: put the files in place
.devin/skills/reference-class-forecasting/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add lyndonkl/claude --skill reference-class-forecasting -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about reference-class-forecasting
How do I install reference-class-forecasting in Claude Code?
Run npx skills add lyndonkl/claude --skill reference-class-forecasting -a claude-code. This installs the file into ~/.claude/skills/reference-class-forecasting/SKILL.md.
What does reference-class-forecasting do?
It provides structured workflows to anchor predictions in empirical frequencies of past events. It checks comparison classes, verifies sample sizes, and models multi-step probability chains.
How large does the reference class need to be?
The framework expects at least 20 to 30 past examples. Anything below 20 requires widening the category because small samples carry heavy uncertainty.