single-agent-loop-kit
CLAUDE.md101 starsChanged 24 days agoApache-2.0View on GitHub
This is the single agent loop kit: a Claude Code skill (skill/), four scripts, a reference loop, and a runnable example. README.md has the map. Follow examples/support-desk/RUNBOOK.md. During the audit, do not open examples/support-desk/ANSWER-KEY.md, examples/support-desk/single_loop/ or examples/support-desk/example-run/. They give away the verdicts the audit is supposed to reach from the code and traces. The skill belongs in their agent repo, not here. Point them at "Use it on your own system" in README.md. - Run python3 -m unittest discover tests…
In this repository2 other files it gives its agents.
- returnsSkill
- single-agent-loopSkill
The file
We split this CLAUDE.md into the rules it gives an agent
3 rules. About 0 are common, 0 are distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 8
Follow examples/support-desk/RUNBOOK.md.
- Very specific: line 20
Run python3 -m unittest discover tests before saying a change works.
- Very specific: line 21
No model ids and no prices in code or docs.
≈ “Common” is an estimate: at least 1 in 5 of the 26,980 CLAUDE.md 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
Working in this repo
This is the single agent loop kit: a Claude Code skill (skill/), four scripts, a reference
loop, and a runnable example. README.md has the map.
If the person wants to try the kit
Follow examples/support-desk/RUNBOOK.md. During the audit, do not open
examples/support-desk/ANSWER-KEY.md, examples/support-desk/single_loop/ or
examples/support-desk/example-run/. They give
away the verdicts the audit is supposed to reach from the code and traces.
If the person wants to use it on their own system
The skill belongs in their agent repo, not here. Point them at "Use it on your own system"
in README.md.
Rules for changes to the kit
- Run
python3 -m unittest discover testsbefore saying a change works. - No model ids and no prices in code or docs. The model comes from
KIT_MODEL, and prices are arguments tocompare.py. The one exception is a record of a run (an example's results and its write-up), which names the public model id the numbers came from. - Anything that calls a real model spends money. Say how big the run is and get a yes first.
KIT_MODEL=fakeruns the example offline for plumbing checks. Its scores mean nothing. - Say what was actually run. A new live run goes in
examples/support-desk/example-run/with its results as they came out, including the bad ones. Keep "What has run on a real model" inREADME.mdtrue, and say in the pull request when a change was only checked offline. - The scripts in
skill/scripts/stay dependency-free, since the skill gets copied into other repos on its own.
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.