court
Skill3 starsChanged 44 days agoMITView on GitHub
Put an idea on trial in front of a jury of 12 Claudes. A prosecutor Claude argues it fails, a defense Claude argues it works, they rebut each other, 12 juror Claudes with different backgrounds vote on their own, and a judge Claude reads the verdict and the conditions that would flip it. Use when the user wants honest pushback instead of agreement, says Claude just agrees with everything, asks "is this a good idea", wants an idea, plan, pitch or claim stress-tested, or says "court" or "put it on trial".
4 other projects use a version of this, 2 of them edited.
Adopted by
Most recently taken up by
- hamzzaqureshi/gravidextheir own version, first seen Nov 8, 2026
- AhmedMerza/claude-skillsedited, first seen Nov 1, 2026
- AhmedMerza/agent-skillsedited, first seen Nov 1, 2026
- MBehtemam/Montagenttheir own version, first seen Sep 24, 2026
Other versions4 projects use a version of this.
Word for word · 0
None found.
Edited · 2
- AhmedMerza/agent-skills★ 286% same
- AhmedMerza/claude-skills★ 286% same
Wrote their own version · 2
- MBehtemam/Montagent★ 3same name
- hamzzaqureshi/gravidex★ 1same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
The file
Show its rules and text
We split this skill into the rules it gives an agent
13 rules. 0 are common, 8 are distinctive, and 5 are very specific to this project.
0
Common
Most of the 4 edited versions we read keep these.
8
Distinctive
This author's own choices. Worth a look.
5
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 29
If that is blank, or still reads like a placeholder, take the case from the conversation: the idea, plan or claim the user most recently …
- Very specific: line 34
Every piece of bookkeeping goes through court.py in this skill's folder:
- Distinctive: line 47
Read the flags out of the request.
- Very specific: line 55
Run COURT plan with the same flags and tell the user in one line how big the trial is, for example "12 jurors, 17 sub-agent calls", then …
- Distinctive: line 72
Do not add facts the user never gave, and do not write your own opinion of the idea into it.
- Very specific: line 83
Always drive it with COURT next.
- Distinctive: line 86
Every phase works the same way:
- Distinctive: line 95
Wait until every agent in the wave has replied before you launch the next wave.
- Very specific: line 96
After the last wave, run COURT next.
- Very specific: line 103
When COURT next says the trial is over:
- Distinctive: line 115
Never change, soften or re-count the verdict.
- Distinctive: line 117
Never add praise the jury did not give.
- Distinctive: line 118
If the user disagrees with the verdict, offer a retrial with the changes from "Conditions for acquittal" written into a new case file.
“Common” is measured: kept by at least 6 in 10 of the 4 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
court
For when Claude agrees with everything you say. Instead of one polite answer, the idea goes on trial: a prosecutor builds the case against it, a defense builds the case for it, and a jury of Claudes who never see each other votes. You are the clerk of the court. You never argue, never vote, and never soften the verdict.
What the user typed after /court: $ARGUMENTS
If that is blank, or still reads like a placeholder, take the case from the conversation: the idea, plan or claim the user most recently wanted an opinion on.
The tool
Every piece of bookkeeping goes through court.py in this skill's folder:
python3 "${CLAUDE_SKILL_DIR}/court.py" <command>Below, COURT means exactly that command. If the path looks unexpanded, use the "Base directory for
this skill" that Claude Code printed at the top of this skill. The state lives in
.court/<run>/state.json in the current directory, and every command after init finds it through
.court/LATEST.
Step 1: size it
Read the flags out of the request. Everything that is not a flag is the case.
| flag | meaning |
--jury N | N jurors, 3 to 12. Default 12. |
--quick | 6 jurors. |
--seed S | Fixes which jurors sit on a smaller jury. Default: random, and recorded. |
Run COURT plan with the same flags and tell the user in one line how big the trial is, for example
"12 jurors, 17 sub-agent calls", then start. If this skill fired on its own (the user never asked for
a trial), ask once first and offer --quick.
Sub-agents write into .court/ in the current directory. In the default permission mode that is one
approval per file. Suggest accept-edits mode (Shift+Tab) for the trial. Do not change the user's
settings yourself.
Step 2: write the case file
Sub-agents cannot see this conversation. The prosecutor, the defense, every juror and the judge
know only what is in the case file, so write .court/case.md to stand on its own:
- First line: the case in one sentence, in the user's own words where you can ("Quit my job to sell candles on Etsy full time").
- Then everything the user told you that matters: numbers, audience, budget, timeline, what they have already tried, constraints.
- Do not add facts the user never gave, and do not write your own opinion of the idea into it. A case file that leans one way decides the trial before it starts.
Step 3: open the trial
COURT init --case-file .court/case.md [--jury N | --quick] [--seed S]Step 4: the trial
Always drive it with COURT next. It reads the files on disk and tells you the next phase and the
exact command. The phases run in this order: opening, rebuttal, jury, judge.
Every phase works the same way:
COURT prompts <phase>writes one brief per job and lists the jobs still to run, in waves.Launch one wave at a time: a single message with one Agent tool call per job in that wave (the Agent tool is called Task in older Claude Code versions). Each call is:
subagent_type:general-purposedescription:court <job id>prompt:Read <brief path> and follow it exactly. It is your whole brief.
Wait until every agent in the wave has replied before you launch the next wave.
- After the last wave, run
COURT next. If an output is missing it sends you back to the same phase, andpromptsthen lists only the missing jobs. Re-run those once. If a juror fails twice, write the single lineNO VOTEinto its file (COURT check jurylists it) and move on. Never write a vote, a statement or a judgment yourself.
Step 5: the verdict
When COURT next says the trial is over:
COURT renderIt counts the votes, writes VERDICT.md into the run folder and prints it. Show the user the verdict
line, the vote table and the judge's sections exactly as written. Then add at most two lines of your
own: the single condition for acquittal you would start with, and the path to the full trial.
Rules for you, the clerk:
- Never change, soften or re-count the verdict. A guilty verdict is useful information, not bad news to cushion.
- Never add praise the jury did not give.
- If the user disagrees with the verdict, offer a retrial with the changes from "Conditions for acquittal" written into a new case file. Do not argue for either side yourself.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

