Skip to content
Back to skills

Uak Loop

ASecurity

Use while implementing a criterion: implement, run Verify, fix, with a budget and stuck detection; for long mechanical fixes it can run unattended with uak loop. Run this only when the user invokes it by name ($uak-loop or /uak-loop); never start it on your own.

  • 20 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added November 12, 2026
ai-agentsgobashdebugging

Security analysis

A100/100

Pro scans all 2 files and shows the line behind each finding

Scanned November 12, 2026

npx -y skills add JoahanMorales/upiixia-agentic-kit --skill uak-loop --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Uak Loop?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Uak Loop
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/joahanmorales-uak-loop/badge)](https://www.skillsdirectory.com/skills/joahanmorales-uak-loop)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: uak-loop
description: "Use while implementing a criterion: implement, run Verify, fix, with a budget and stuck detection; for long mechanical fixes it can run unattended with uak loop. Run this only when the user invokes it by name ($uak-loop or /uak-loop); never start it on your own."
disable-model-invocation: true
---
<!-- Generated by scripts/gen_harness.py from .claude/commands/uak-loop.md; edit that file, not this one. -->
Arguments (what the user typed after the command, `<arguments>` below): [Verify command]
Verifier: the arguments you were given (none: the task's `Verify`). Loop rules (`.uak/docs/LOOPS.md`):
1. Make the smallest change toward the criterion. For new behavior, the test comes first and fails on its assertion (skill `test-first`, `uak red ID`).
2. `bash .uak/bin/q <verifier>`. If OK, go to step 5.
3. On failure, read only the tail `q` printed. Find the root cause before editing (skill `systematic-debugging`). Never weaken, skip or delete a test or the verifier to make it pass.
4. Repeat, up to 5 attempts. If the same error appears twice, consult the advisor (or `/uak-advise stuck`) ONCE and apply its single experiment. If it fails again, STOP: `bash .uak/bin/uak log ID --state BLOCKED "<exact error>"` (ID = your task), or run `uak decision ID --option "TEXT" --why "TEXT" --reversible yes --authorized yes --class routine` (the reversible fallback), and tell the human in 1 line.
5. Before claiming it works (skill `verification-before-completion`), re-run the verifier and smoke, then commit and push. `bash .uak/bin/uak checkpoint ...` with the next criterion.
Long mechanical loops (lint, types, a flaky failure with a clear error) can run unattended: `bash .uak/bin/uak loop --verify "<cmd>" --agent "claude -p --permission-mode acceptEdits" --max 5 --task ID`.

Files in this skill

  • SKILL.md1.8 KB
  • agents/openai.yaml130 B

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

View all in ai-agents →

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.