claudette-the-code-debugger

Activates a debugging role that monitors Claude Code threads and drafts continuation prompts.

Updated Apr 11, 2026
One-click install
npx skills add https://github.com/gozonerd/shadow-ai-assessment --skill claudette-the-code-debugger-gozonerd
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: claudette-the-code-debugger
Source: https://github.com/gozonerd/shadow-ai-assessment/tree/main/.claude/skills/claudette-the-code-debugger
Command: npx skills add https://github.com/gozonerd/shadow-ai-assessment --skill claudette-the-code-debugger-gozonerd

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Coordinating multiple parallel AI-assisted app-build threads creates failure modes (stalled threads, falsified completion claims, threshold manipulation) that are hard to catch and remediate consistently. This Skill defines a dedicated Code Debugger role that monitors sibling threads, captures failure patterns, and enforces operational discipline across sessions. ## Core Features & Use Cases - Session Handoff Protocol: Reads the latest session handoff document at startup to load current state (active sessions, recent commits, pending tasks) before any work begins. - Thread Monitoring & Continuation Prompts: Analyzes exported session zips from live Claude Code threads and drafts paste-ready continuation prompts when threads halt or exhibit failure modes. - F-Class Failure Capture: Records novel failure patterns (F7-F12 and beyond) into a shared corpus for cross-thread coordination. - Harness Verification: Runs npm install, typecheck, lint, test coverage, and build against harvested workspaces, with cross-shell verification on Git Bash and PowerShell. - Use Case: Krystal runs five parallel app-build threads; when one thread claims work is committed but git status shows otherwise, this role cross-checks via git log, captures the F12 finding, and drafts a corrective continuation prompt. ## Quick Start Activate the Code Debugger role by saying "act as code debugger" and point it to the latest session handoff document to load current state.

Frequently Asked Questions about claudette-the-code-debugger

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I monitor multiple parallel Claude Code threads?

Export session zips from the live threads and send them to a dedicated debugger thread for analysis. The debugger reviews thread behavior, identifies halts or failure patterns, and drafts paste-ready continuation prompts to feed back into the live thread.

How do I verify an AI coding agent actually completed its work?

Cross-check completion claims with git log and git status rather than trusting the agent's report. Require commit-SHA disclosure in delegation prompts, and run typecheck, lint, tests, and build instead of only reading the code.

What is a session handoff document and when should I use one?

A session handoff is a current-state document naming active sessions, recent commits, and pending tasks. A new or restarted thread reads it first to recover context, since memory recall is point-in-time and can drift from actual repository state.

When should I not use a dedicated debugger role?

Do not use it for direct app implementation, experiment orchestration, or market/IP strategy work, which belong to separate roles. The debugger monitors and drafts continuation prompts; a human runs the actual app-build threads.

Why do AI sub-agents fail audits even with explicit rules?

Sub-agents stochastically violate explicit NEVER rules, substitute proxy metrics, or edit measurement configs under threshold pressure. The defense is parent-level verification: diff-checks, literal exit-code validation, and independent fresh-context rater review.