Skip to content
Back to skills

Debug Loop

ASecurity

Build a tight failing feedback loop before diagnosing hard bugs, flaky tests, regressions, or performance issues.

  • 3 stars
  • 0 votes
  • 0 copies
  • 1 view
  • Added September 5, 2026
ai-agentsperformance

Works with

  • cli

Security analysis

A100/100

Scanned September 5, 2026

npx -y skills add me-cedric/agent-compass --skill debug-loop --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Debug Loop?

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

Security grade badge for Debug Loop
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/me-cedric-debug-loop/badge)](https://www.skillsdirectory.com/skills/me-cedric-debug-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: debug-loop
description: >
  Build a tight failing feedback loop before diagnosing hard bugs, flaky tests,
  regressions, or performance issues.
risk_level: medium
writes_files: false
requires_tools: []
---

# Debug Loop

Use when bug is unclear, flaky, slow, or reported only as "broken".

## Rule

No theory before loop. First create one command that can catch the exact symptom.

## Loop Ladder

Stop at first rung that catches the bug:

1. Existing focused test.
2. New smallest regression test.
3. CLI or script with fixture input.
4. HTTP request against running service.
5. Playwright/browser script for UI.
6. Captured trace replay.
7. Tiny harness around one function/module.
8. Stress loop for flaky bugs.

## Tighten

Loop must be:

- **Specific**: asserts user's symptom, not "does not crash".
- **Fast**: seconds if possible.
- **Deterministic**: same verdict, or high reproduction rate for flakes.
- **Agent-runnable**: no manual clicks unless wrapped in a written HITL script.

## Work

1. Name loop command and run it once.
2. Minimize inputs until every remaining part is load-bearing.
3. Change one thing at a time.
4. Keep the loop as the regression check.
5. Report command, red/green evidence, and remaining uncertainty.

If no loop can be built, stop and ask for the missing artifact: logs, HAR,
fixture, environment access, or permission for temporary instrumentation.

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.

Comments

Loading comments…