Cozempic Doctor

When Claude Code starts acting weird,hooks silently failing, subagents throwing 403s, or sessions hanging on Windows,this is your first stop. It runs diagnostics for ten specific failure modes including the infamous trust dialog hang bug, corrupted tool_use blocks that cause 400 errors, and orphaned tool results from concurrent writes. The --fix flag attempts automatic remediation. It's essentially a specialized linter for Claude Code's internal state, catching the kinds of gnarly edge cases that happen when an AI coding assistant manages its own configuration files and spawns child processes. Beats manually grepping through .claude.json trying to figure out why everything suddenly broke.

GitHub stars: 332
last updated: Jun 11, 2026

View on GitHub

Install

$ npx -y skills add ruya-ai/cozempic --skill cozempic-doctor --agent claude-code

Installs into .claude/skills of the current project.

Files

Run cozempic health checks:

cozempic doctor

Checks for:

Trust dialog hang — Windows resume bug where .claude.json trust entry causes hangs
Hooks trust flag — missing hasTrustDialogHooksAccepted causing hooks to silently fail (v2.1.51+ bug)
Agent model mismatch — spawned subagents not inheriting team lead's model causing 403 errors
Claude.json corruption — truncated or invalid .claude.json from concurrent write races
Corrupted tool_use — tool blocks with names >200 chars causing 400 API errors
Orphaned tool_results — tool_result blocks missing matching tool_use causing 400 API errors
Zombie teams — stale agent team directories accumulating on disk
Oversized sessions — sessions that may trigger compaction issues
Stale backups — old .bak files consuming disk space
Disk usage — total Claude Code storage footprint