Context profiles are JSON files that tell Claude (or any AI) how to respond — voice, depth, citation rules, response shape. Load a profile once, every subsequent prompt inherits it. Senior dev, copywriter, analyst, strategist — each gives you a different lens on the same input.
Drop these JSON files into Claude Code's context (via /profile load skill or paste into your system prompt). Every subsequent response shifts.
Strips passive voice. Cuts every hedge ("perhaps," "might," "could"). Leads with the answer, then the evidence. Treats every paragraph as a candidate to delete.
Output reads like a senior copywriter — confident, specific, conversion-aware. Will refuse to write fluff even when asked.
Pushes back on overengineering. Names trade-offs out loud. Won't add abstractions unless they solve a real, current problem. Will tell you when the "right" answer doesn't matter for your scale.
Writes code like someone who's been on-call. Defensive, observable, easy to debug at 3am.
Each profile follows the same shape. Two new profiles drop monthly. Community members can submit their own — best ones get added.