spar
Skill1 starsChanged 9 days agoMITView on GitHub
Use when the user wants a plan, design or decision stress-tested, or asks to spar or to be grilled. Interviews in rounds until no decision is open.
3 other projects use a version of this.
Adopted by
Most recently taken up by
- jared833/claude-code-hookstheir own version, first seen Oct 21, 2026
- lossyrob/skillstheir own version, first seen Aug 27, 2026
- DanaResearchGroup/agent-skillstheir own version, first seen Aug 26, 2026
Other versions3 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 3
- lossyrob/skills★ 4same name
- jared833/claude-code-hooks★ 2same name
- DanaResearchGroup/agent-skills★ 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.
In this repository13 other files it gives its agents.
- architectureSkill
- artifactsSkill
- buildSkill
- debtSkill
- deep-reviewSkill
- diagnoseSkill
- distributed-architectureSkill
- finalizeSkill
- handoffSkill
- rulesSkill
- security-checkSkill
- specSkill
- whichSkill
The file
We split this skill into the rules it gives an agent
5 rules. 0 are common, 3 are distinctive, and 2 are very specific to this project.
0
Common
Most of the 3 edited versions we read keep these.
3
Distinctive
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 13
Read the request and the code it touches.
- Very specific: line 14
Ask every ready decision in one round.
- Distinctive: line 21
Read the answers and mark those decisions settled.
- Distinctive: line 22
When no decision is open, list the decisions, one line each.
- Very specific: line 31
Ask in your message, not with AskUserQuestion.
“Common” is measured: kept by at least 6 in 10 of the 3 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
Spar
Interview the user until you both agree on every decision that changes what gets built or how it is checked.
Rounds
- Read the request and the code it touches. List the open decisions. Some decisions depend on others. A decision is ready when every decision it depends on is settled.
Ask every ready decision in one round. Number the questions. Give your recommended answer for each, worded so that
yesaccepts it. Use this format:Q1.
<the question>? <the context, and the options when there are options> Recommended: <your answer, with a one-line reason>- End the round with this line: "Reply
yesto take every recommendation, or answer by number." Then stop and wait. - Read the answers and mark those decisions settled. Find the decisions that are ready now, and ask the next round. A question that depends on an answer from the current round waits for the next round.
- When no decision is open, list the decisions, one line each. Ask the user to confirm the list, and do not act on it before the user confirms.
- For a confirmed decision that is hard to reverse and beat a real alternative, offer to record it as an ADR in
docs/adr/, in the format ofarchitecturestep 6. Write it only on yes. Say nothing about ADRs for other decisions.
Rules
- Facts are your job. Look them up in the code, the docs or the tools. Give a slow search to a subagent when you can, and meanwhile ask the questions that do not depend on it. Never ask the user for a fact you can find.
- Decisions belong to the user: goals, users, scope, constraints and what done looks like. Decide internal details yourself.
- Each question must change the build or the check. Ask no filler.
- A clear request has no open decisions. Then say so and ask nothing.
- Ask in your message, not with
AskUserQuestion. Fixed options push the user to pick instead of answer, and the tool is missing in headless runs. - There is no limit on rounds. The user can reply
wrap upat any time. Then stop asking, and list each open decision with your recommended answer as an assumption.
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.