closed-loop
Skill1 starsChanged 5 months agoMITView on GitHub
Run a bounded 5-stage agent loop (discover → plan → execute → evaluate → improve) against a defined goal, rubric, and stop rule. Use for marketing production, content ops, CRO, and any repeatable asset workflow.
2 other projects use a version of this.
Adopted by
Most recently taken up by
- SHYXIN/skillstheir own version, first seen Nov 1, 2026
- huytieu/COG-second-braintheir own version, first seen Jul 30, 2026
Other versions2 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 2
- huytieu/COG-second-brain★ 1.3ksame name
- SHYXIN/skills★ 6same 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.
- adversarial-verifySkill
- cluster-feedbackSkill
- doubt-gateSkill
- fleet-loopSkill
- hill-climbSkill
- keep-greenSkill
- loop-auditSkill
- loop-readinessSkill
- loop-stateSkill
- prod-watchSkill
- scaffold-planningSkill
- ship-preflightSkill
- source-groundedSkill
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 1 is distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 18
Not for open-ended exploration — use fleet-loop for that.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Closed Loop
A bounded looping harness for producing high-quality assets. You define the goal, inputs, and rubric. The loop runs discover → plan → execute → evaluate → improve until every gate passes or the iteration cap is hit.
When to use
- Landing page rewrites
- Ad creative generation
- Email nurture sequences
- Content repurposing
- SEO articles
- Offer research and positioning
Not for open-ended exploration — use fleet-loop for that.
The 5 stages
| Stage | Job |
| Discover | Research the market, buyer, offer, competitors, source material, or objections |
| Plan | Map the work into clear steps with acceptance criteria |
| Execute | Draft, build, or produce the asset |
| Evaluate | Score the output against the rubric; identify what's below threshold |
| Improve | Fix weak sections, re-run evaluation, repeat |
Reusable prompt template
Goal:
[Business outcome — not just the asset type.]
Inputs:
- Audience:
- Offer:
- Source material:
- Brand voice:
- Constraints:
Loop:
1. Discover: research relevant context, extract useful facts.
2. Plan: break the work into clear steps.
3. Execute: produce the first draft.
4. Evaluate: score against the rubric below.
5. Improve: fix anything below the pass threshold.
6. Repeat until every gate passes or you hit the stop rule.
Eval rubric:
- Clear target buyer: pass/fail
- Specific pain or desire addressed: 1–10
- Strong proof (evidence, data, testimony): 1–10
- Differentiated angle (not generic): 1–10
- Concrete examples: 1–10
- No generic filler: pass/fail
- CTA or next step is obvious: pass/fail
Stop rule:
Stop when every scored item is 8/10 or higher and every pass/fail gate passes.
If the loop runs 3 times without improvement, stop and explain the blocker.Worked examples
Landing page
- Discover: ICP, offer, competitors, objections, current page
- Plan: hero, proof, mechanism, offer, CTA, FAQ
- Execute: rewrite each section
- Eval: clarity, specificity, proof, objection handling, CTA strength
- Improve: fix any section under 8/10
Ad creative
- Discover: customer language, reviews, competitor ads, pain points
- Plan: angles (pain, proof, contrarian, urgency, identity, demo)
- Execute: hooks, body copy, CTA
- Eval: reject generic claims, weak hooks, unclear offer
- Improve: 3 stronger variants per winning angle
Content repurposing
- Discover: thesis, claims, proof, sharp lines from source
- Plan: formats (LinkedIn, X thread, newsletter, short script)
- Execute: draft each asset
- Eval: voice fidelity, platform fit, specificity, usefulness
- Improve: cut generic sections, add concrete examples
Slop prevention
- The eval gate is the difference between compounding quality and compounding slop.
- Where an objective gate exists (test, build, lint, word count, link check), it outranks the rubric score — a rubric without any objective signal is the agent grading its own homework.
- Loose goals ("make it better") always produce generic output. Specify what "better" means.
- Cap iterations. If quality stops improving after 3 runs, surface the blocker — don't iterate into noise.
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.