Improve how you work with AI. Portable Claude Code and Codex skill with evidence-based assessment and evaluation.
# Add to your Claude Code skills
git clone https://github.com/intelligenceagents/better-loop-skillGuides for using ide extensions skills like better-loop-skill.
better-loop-skill is an open-source ide extensions skill for AI coding assistants such as Claude Code, Codex CLI, and ChatGPT, built by intelligenceagents. Improve how you work with AI. Portable Claude Code and Codex skill with evidence-based assessment and evaluation. It has 0 GitHub stars.
better-loop-skill's catalog security scan is still queued. You can run an instant dependency and prompt-injection check now with the "Scan for vulnerabilities" button above.
Clone the repository with "git clone https://github.com/intelligenceagents/better-loop-skill" and add it to your Claude Code skills directory (see the Installation section above).
better-loop-skill is primarily written in TypeScript. It is open-source under intelligenceagents on GitHub, so you can review or fork the full source.
Yes. SkillsLLM lists many other IDE Extensions skills you can browse and compare side by side. Open the IDE Extensions category from the badge at the top of this page, or use the Related Skills and comparison links further down to weigh better-loop-skill against similar tools.
No comments yet. Be the first to share your thoughts!
Unlocks once the catalog security scan passes (runs nightly).
⚠️ Third-Party Software Notice
This skill is third-party open-source software developed and hosted independently on GitHub. SkillsLLM is an informational directory and does not control or maintain the underlying repository.
Any security checks, ratings, or warnings displayed by SkillsLLM are automated and limited in scope. They do not constitute a security certification or guarantee that the software is safe, error-free, or free from malicious code, vulnerabilities, compromised dependencies, or prompt-injection risks.
Review the source code, permissions, dependencies, and configuration before installing or running any third-party skill. Use is at your own risk. To the maximum extent permitted by applicable law, SkillsLLM is not liable for losses arising from third-party software.
The deep catalog scan for this skill is still queued. Run an instant dependency check now instead.
Get better at working with AI. One real task at a time.
Better Loop is a private coaching skill for Claude Code and Codex. Bring a task you worked on with AI. Leave with a clearer prompt, one change to try, and a concrete way to check the next result.
You choose the work it may read. Your history stays on your laptop. No Better Loop account is needed; sharing is optional.
Get started · How it works · Read the skill
Install from source on main. These early packages are not published on npm. You need Git, Node 22 (the selected patch is in .nvmrc), npm 10.9.0, and a configured Claude Code or Codex installation.
git clone --branch main --single-branch https://github.com/intelligenceagents/better-loop-skill.git
cd better-loop-skill
npm ci
npm run build:local
git rev-parse HEAD
node skills/better-loop/scripts/detect-helper.mjs
Keep the printed commit with your installation details. Follow the host install guide to copy the skill into your chosen project and connect it to this helper. It covers existing installations, updates and rollback.
In that project, invoke /better-loop in Claude Code or $better-loop in Codex, then ask:
Review my current repository only. Help me improve how I work with AI: suggest one prompt change and one check for my next attempt. Ask for any missing task or dedicated local checkpoint path. Save the result there and keep it private.
For several repositories, replace the first sentence with: “Review only these repositories: [your exact paths].” There is no background history scan.
Building downloads public dependencies. Host reasoning uses your configured model provider, which may process the local material you select.
Help me clarify my audience, output format, collaboration style, feedback, delegation boundaries, verification habits and next acceptance check. Propose a better prompt and a working agreement for this project. Show me the exact diff before applying it.
Those seven choices guide a project-scoped AGENTS.md for Codex or CLAUDE.md for Claude Code. You can explicitly select different host/model profiles. Review and approve the exact proposal; keep its rollback.
This adapts instructions, not model weights. Instruction following varies; a saved preference is preparation, and the next checked task supplies evidence. See the personal coach guide.
Return to my approved checkpoint at [exact local path]. Review only changed selected evidence, recall the earlier recommendation, and check whether the later result supports improvement. Keep missing measurements explicit.
The same selected checkpoint preserves history across chats and both hosts. Unchanged work creates no new assessment or progress credit; changed scope or conditions can prevent comparison. Open a private progress viewer or use the host request guide.
| Work you do | Something to practice |
|---|---|
| Coding | Define a failure and its check before changing code |
| Analysis and finance | Make assumptions and discrepancies explicit |
| Research and strategy | Separate supported conclusions from uncertainty |
| Mathematics and science | Test assumptions, units and counterexamples |
| Writing and design | Specify the audience and review criteria |
| Operations and education | Clarify dependencies and acceptance checks |
| General tasks | Define a useful result before delegating |
Try one of eight shared practice briefs. These are starting points; host support does not establish assessment validity in every field.
The public beta website supports email sign-in, reviewed publishing and owner version history. You can unpublish or delete what you shared. Private coaching still needs no account.
Optional sharing prepares a minimized improvement story. Raw code, documents, workflows and transcripts are not Better Loop uploads. Export requires privacy review and your approval of the exact content and purposes; changing either invalidates approval. See privacy and reviewer setup.
Private coaching and history remain useful without sharing.
Walk through a complete loop, from the skill to a public story.
Behavioral coaching draws on Anthropic's 4D framework and 11 observable indicators. Outcomes, efficiency, missing evidence and verification stay separate. Better Loop has no universal person score or validated hiring prediction.
These are bounded engineering observations, not proof of human learning or guaranteed savings. Current status records implementation and verification limits.
MIT licensed, including commercial and corporate reuse. Integration notifications are voluntary and should contain no confidential information.
After the source setup above, add Python 3.11 or later:
python3 -m venv .venv
source .venv/bin/activate
python -m pip install -r requirements-dev.txt
npm run check
These checks use local fixtures and make no model calls. See development for browser and offline package checks.