loop
Skill8 starsChanged 4 months agoMITView on GitHub
The build loop - locate yourself, investigate the real code, build one slice completely, verify it for real, fix and re-verify, then close and hand off. Use for any session that is implementing something from the plan.
55 other projects use a version of this.
Adopted by
Most recently taken up by
- Eclipter/agentstheir own version, first seen Nov 8, 2026
- SethGammon/Citadeltheir own version, first seen Nov 4, 2026
- drevendev/HungryCrabtheir own version, first seen Nov 1, 2026
- Railly/skillstheir own version, first seen Nov 1, 2026
- HeshamFS/swarmweavertheir own version, first seen Nov 1, 2026
- jtaroreh/agystacktheir own version, first seen Oct 29, 2026
Used with
Projects with it also use review-bugbot, origin and 10 more.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
0mlkrizzz655335vSkillreview-bugbot
缺陷导向代码审查。用 Bugbot 子代理找 bug。
Also in 4 projects that use this
0mlkrizzz655335vSkillorigin
origin CLI 管理。安装、登录、更新、修复 origin.cursor.com 的 CLI。"disabled-environments:
Also in 3 projects that use this
PandelisZSkillupdate-cli-config
View and modify Cursor CLI configuration settings in ~/.cursor/cli-config.json. Use when the user wants to change CLI settings, configure permissions, switch approval mode, enable vim mode, toggle display options, configure sandbox, or manage any CLI preferences.
Also in 4 projects that use this
32103210Skillmigrate-to-skills
Convert 'Applied intelligently' Cursor rules (.cursor/rules/*.mdc) and slash commands (.cursor/commands/*.md) to Agent Skills format (.cursor/skills/). Use when you want to migrate rules or commands to skills, convert .mdc rules to SKILL.md format, or consolidate commands into the skills directory.
Also in 4 projects that use this
32103210Skillupdate-cursor-settings
Modify Cursor/VSCode user settings in settings.json. Use when you want to change editor settings, preferences, configuration, themes, font size, tab size, format on save, auto save, keybindings, or any settings.json values.
Also in 4 projects that use this
MicrosoftSkillautomate
Propose an automation from a completed task or a new idea, then create it after approval. Use when the user asks to create an automation or wants scheduled local or cloud agent work.
Also in 4 projects that use this
- +3 more from 0mlkrizzz655335v
- +1 more from 32103210
Other versions55 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 55
- SethGammon/Citadel★ 927same name
- breezewish/CodexPotter★ 629same name
- jtaroreh/agystack★ 128same name
- dannys-code-corner/incan★ 73same name
- ww-w-ai/bkit-gemini★ 66same name
- IsuruMaduranga/one-code★ 50same name
- unclejobs-ai/second-claude-code★ 41same name
- BULDEE/ai-craftsman-superpowers★ 35same name
- Hassaan146/claude-skills★ 18same name
- woxiangyangzhimao/skills★ 14same name
45 more not listed
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 repository6 other files it gives its agents.
- loop-closeSkill
- loop-councilSkill
- loop-doctorSkill
- loop-planSkill
- loop-setupSkill
- loop-verifySkill
The file
We split this skill into the rules it gives an agent
9 rules. 0 are common, 2 are distinctive, and 7 are very specific to this project.
0
Common
Most of the 50 edited versions we read keep these.
2
Distinctive
This author's own choices. Worth a look.
7
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- could be automatic: a setting could enforce it
- Distinctive: line 12
Run these six steps for every item.
- Very specific: line 16
Read .loop/STATE.md (what is done, what is next, what is parked) and .loop/GOTCHAS.md (traps this project has already sprung).
- Very specific: line 31
If a question is genuinely open — or the item touches money, permissions, auth, or schema — run the loop-council skill *before* writing …
- Very specific: line 43
Run the slots in .loop/VERIFY.md that apply.
- Distinctive: line 55
Commit on the working branch, staging only your own files (never git add -A when someone else has uncommitted work).
- Very specific: line 57
Update .loop/STATE.md: mark the item [x] and add one line — *what you verified and how*.
- Very specific: line 59
If you were blocked, write PARKED: <why> + exactly what is missing and move to the next item.
- Very specific: line 61
Add to .loop/DECISIONS.md if you chose between real alternatives, and to .loop/GOTCHAS.md if something fooled you — the next session should …
- Very specific: line 68
Only for what only they can unblock: a key or an access you do not have, moving real money, a business/legal/pricing decision, or …
“Common” is measured: kept by at least 6 in 10 of the 50 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
The loop
Nothing is done until it has been observed working. A green build, "I read the code and it looks right", and "this should work" are not verification. The only valid signal is observed behaviour.
Run these six steps for every item. Do not skip 0 and do not skip 5.
0 · Locate yourself
- Read
.loop/STATE.md(what is done, what is next, what is parked) and.loop/GOTCHAS.md(traps this project has already sprung). git log --oneline -15andgit status. Anything already committed is not rebuilt. A lot of what a document calls missing is already built and just not visible — confirm against the code before redoing it.- Uncommitted files that are not yours may belong to a parallel session. Do not
touch them, and do not reuse a shared counter they may have taken (see the shared
resource in
.loop/VERIFY.md— read its true state with the command declared there). - The next item is the first unchecked
[ ]in.loop/STATE.md. You do not need new instructions to continue.
1 · Investigate
Open the real files you are about to change. Map what exists, what is editable,
what is missing. Never work from memory of an API or a schema.
If a question is genuinely open — or the item touches money, permissions, auth, or
schema — run the loop-council skill before writing code, not after.
2 · Build, completely
One item at a time, finished. If completing it requires one more thing, that thing gets built too: it is not reported as a leftover and it is not offered as a question. An item is complete only if it would be safe to leave exactly as it is, forever — no dead end, no control that does not do what it says, no setting with no effect.
3 · Verify for real
Run the slots in .loop/VERIFY.md that apply. The rule for each: it must produce an
artifact or an observation — a rendered page, a row, a file, a screenshot, a log
line that only exists if the new code ran. Forbidden as evidence: an exit code, an
HTTP 200, and "the build passed". Those prove the code compiled, not that it works.
4 · If it fails, fix it and verify again
Repeat 2 → 3 → 4 until it actually works. This is the loop. Do not move on with something unverified, and do not describe a failure as a partial success.
5 · Close
- Commit on the working branch, staging only your own files (never
git add -Awhen someone else has uncommitted work). Never push to the protected branch. - Update
.loop/STATE.md: mark the item[x]and add one line — what you verified and how. Record anything you could not verify asGAP: <why>. - If you were blocked, write
PARKED: <why> + exactly what is missingand move to the next item. Park, never abandon, and never get stuck on one item. - Add to
.loop/DECISIONS.mdif you chose between real alternatives, and to.loop/GOTCHAS.mdif something fooled you — the next session should not fall for it. - Report the delta honestly: what is verified, what is a GAP, what is blocked on the owner. Then continue with the next item. Do not ask "am I done, shall I continue?".
When to actually ask the owner
Only for what only they can unblock: a key or an access you do not have, moving real money, a business/legal/pricing decision, or publishing. Everything technical you decide yourself. "Do you want me to test it?" is never a valid question — if it matters, you already tested it.
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.

