Decision Protocol
A /decide command that sizes every business decision before analysing it: a $30 tool gets a fast pass, a $20K hire gets four specialist agents arguing about it.
My verdict: Every meaningful business decision I have made since April runs through this. The record folder it builds has become the most honest document I own.
What it is
A Claude Code command that refuses to answer “should I buy X” the way a chat window would. It triages first: money and hours decide the tier (micro, small, medium, large), reversibility decides the caution level, and only then does analysis run, at matching depth. Micro decisions resolve in seconds. Large ones dispatch four specialist sub-agents (Strategy, Critic, Finance, COO), make them peer-review each other, and tell you which human expert to consult before committing.
Two habits are built into every run:
- The reframe. “Should I upgrade my computer” becomes “what removes friction from revenue in the next 90 days.” The surface question is almost never the real one.
- The record. Every decision writes a dated file with the call, the confidence percentage, and a prediction. At the review date you fill in what actually happened. That loop, over months, is how your judgment improves.
Who it is for
Anyone running a small business alone, where every decision is yours and the failure mode is either agonizing over $50 tools or impulse-committing to $5K programs. The tier system gives each its proper weight.
The part nobody else has
Emotional pattern detection. Each run scans for eight distortions (sunk cost, scarcity theater, validation seeking, shiny object among them), names the one it sees with evidence, and tags the record. My own folder shows which bias I repeat. Yours will too.
Setup outline
- Copy one file into
~/.claude/commands/. - Optionally point the config block at your goal and plan files so decisions are analysed against your actual current focus.
- Run
/decide "your question".