- Home
- Skills
- Agents & Orchestration
- Loop Stop
Works with the AI Agents you already use
Loop Stop
- **Before launching:** run heavy tasks in batches and in the foreground, and note any measured baseline to compare against.
Free
Loop Stop
Example session with this skill installed
I'm going to launch a subagent to fix all the failing tests in tests/ overnight. Brief it using the loop-stop skill.
- Read your context and instructions
- Compiled the loop stop
- Generated the document
Example brief. Written by following the skill's rules, to show the shape of the result; it is an illustration, not a recorded run.
Goal and why: get the failing tests in tests/ passing so tomorrow's release branch is green. Knowing the purpose lets the subagent skip tests that are meant to fail.
May touch: files under src/ and tests/ only. Nothing in config/ or .github/.
Budget: 15 turns or 20 minutes, whichever comes first.
Baseline: 12 failing tests at launch (the test output is saved to baseline.txt).
Stop and rethink (don't raise the limit and don't retry the same way) if:
- Two test runs in a row show the same failures with the same names.
- The same error appears twice: change the cause before trying again.
- The budget above is used up.
- A git conflict or a locked file blocks progress: report it, don't force it.
Result handling: at most 3 follow-up rounds, each with its reason. On round 3 the task is closed as APPLIED or DISCARDED, with a one-line verdict.
Next steps
- Launch it in the foreground, not in the background.
- Check the failure count after the first five turns.
- Compare with
baseline.txtbefore you accept the result.
loop-stop.pdf
PDF · document
Example file from a real run - the skill writes it into your workspace.
Connects securely to your tools. The creator never sees your data.
About this skill
What it is
Loop Stop is one short SKILL.md: a checklist your agent follows when you launch a subagent, a sweep or any repeating task, and when you watch one run.
What it tells the agent to do
- When briefing: write the goal and why, the files it may touch, a budget (turns, tokens or minutes) and the stop rules.
- Stop and rethink (don't raise the limit or retry the same way) if there are two checks in a row with no measurable progress, the same error twice, the budget is exceeded, or a git conflict or locked file blocks progress.
- When the result comes back: at most three follow-up rounds, each with a reason. The third closes the task as APPLIED or DISCARDED.
- Before launching: run heavy tasks in batches and in the foreground, and note any measured baseline to compare against.
Why we wrote it
We watched agents sit waiting on their own process, or repeat the same attempt again and again, burning quota. This is the rule set we now use to stop that.
Good to know
- It is written instructions for an AI, not a program. It can't force an agent to obey; it makes the right behaviour easy to ask for and easy to check.
- It doesn't monitor anything by itself: you or the parent agent apply the stop rules.
- Written and used in Claude Code. It follows the open SKILL.md format, but we haven't run it in other tools.
- Plain text, MIT licence. It is a few lines: read it before you install it.
More free skills and guides: https://happy-program-zero.github.io
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Agents & Orchestration.
It is written instructions, not a program: it can't force an agent to obey. It has no monitoring script, so you or the parent agent apply the stop rules. Written and used in Claude Code; not run in other tools.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Free to use
No account required to browse
Trust & safety
Security scan
Security scan passed 18 days ago
- Free with an Agensi account
Permissions
No special permissions declared or detected