/jsonReturns a valid JSON object only — no surrounding text, no explanation.
Category: Output Formatting
Before / After
Give me the data as JSON.
/json Give me 5 example users with name, email, and signup date.
What changes in Claude's response
You get exactly one JSON object — no preamble, no explanation, no markdown code fence. You can pipe the output directly into `jq` or paste it into your code. The schema is consistent (Claude won't randomly rename keys between calls), which makes /json safe for scripting.
Full deep dive locked
The complete analysis for /json — including when NOT to use it, the failure modes, and the codes it stacks with — is in the full cheat sheet along with 119 other tested prefixes.
Lifetime updates · 7-day refund · 3 tiers
More Output Formatting codes
ARTIFACTSStructures the output as discrete numbered deliverables you can act on independently.
/tableForces a markdown table for comparable data — even when Claude wouldn't pick a table on its own.
/bulletsPure bulleted output, zero prose. Even the introduction is a bullet.
/tldrForces a one-line summary above the full answer so you can skim or read.
/yamlSame as /json, but YAML — for configs, GitHub Actions, Kubernetes manifests.
Get new codes + workflows weekly
Join 30+ developers getting tested prompt codes, skill files, and Claude workflows. One email when there’s something worth sharing.