doctor · rohitg00 — Claude Code command: configuration diagnostic
doctor scans your pro-workflow and Claude Code setup: checks hooks, database, context usage, MCP servers, and git state. Spots broken hook scripts and settings conflicts before they cause problems.
MIT ★ 2,966 updated 2026-09-29
/plugin marketplace add rohitg00/pro-workflow
/plugin install pro-workflow@pro-workflowWhat doctor does
The doctor command runs a full diagnostic suite. It verifies that pro-workflow commands are in place, the database is accessible, and hooks are active and working. Each hook—commit-validate, secret-scan, git-blast-radius—gets tested on both valid and invalid inputs; all should return OK.
It also measures context usage (alert at 70–90%, critical above 90%), counts CLAUDE.md lines (split if over 150), tallies active MCP servers (aim for under 10), checks git status and uncommitted changes, and flags conflicts between project-level and user-level settings.json. Wiki and council status show without exposing environment values.
Finally, it outputs a summary: installation status, hook state, context percentage, MCP count, and CLAUDE.md size. Any failure gets a quick-fix table—from copying commands to disabling unused servers.
What it helps with
- Verify hook scripts are active and blocking unsafe git and secret operations
- Find conflicts between project and user settings before they interfere
- Monitor context and CLAUDE.md size ahead of heavy sessions
- Confirm pro-workflow installation and database availability
- Troubleshoot inactive hooks or incorrect script paths
Example prompts for the agent
- Run
/doctorbefore critical commits to ensure hooks catch dangerous patterns - Use
/doctorif context feels sluggish or Claude behaves unexpectedly - Call
/doctorafter updating pro-workflow to verify nothing broke
How to install doctor
1Plugin marketplace
/plugin marketplace add rohitg00/pro-workflow/plugin install pro-workflow@pro-workflow2Manually: put the files in place
.claude/commands/pro-workflow-doctor.mdThe built-in /doctor overrides a file with the same name, so the file is renamed. Call it with /pro-workflow-doctor. For all projects use ~/.claude/commands/
Manually: put the files in place
~/.codex/prompts/doctor.mdCall it with /prompts:doctor. Codex marks this format deprecated in favour of skills
Manually: put the files in place
.github/prompts/doctor.prompt.mdManually: put the files in place
.gemini/commands/doctor.tomlGemini stores commands as TOML: the prompt text goes into the prompt field
Questions about doctor
How do I install doctor in Claude Code?
Run /plugin marketplace add rohitg00/pro-workflow, then /plugin install pro-workflow@pro-workflow, and point to .claude/commands/pro-workflow-doctor.md.
What does doctor do?
It audits pro-workflow installation, hook status (commit-validate, secret-scan, git-blast-radius), context and CLAUDE.md size, active MCP servers, git state, and settings conflicts. Produces a summary and suggests fixes.
What if doctor says 'Hooks: INACTIVE'?
Hooks aren't firing. Check that hooks.json is in the right location, script paths use ${CLAUDE_PLUGIN_ROOT} or absolute paths, and scripts have execute permissions.
Source text
doctor.md · MIT
---
description: Diagnose pro-workflow setup and Claude Code configuration
---
# /doctor - Configuration Health Check
Run a diagnostic on your pro-workflow and Claude Code setup.
## Check These
### 1. Pro-Workflow Installation
```bash
ls -la ~/.claude/commands/wrap-up.md 2>/dev/null && echo "Commands: OK" || echo "Commands: MISSING"
ls -la ~/.pro-workflow/data.db 2>/dev/null && echo "Database: OK" || echo "Database: MISSING"
```
### 2. Hooks Status
Check if hooks.json is loaded and hooks are active. Run:
```bash
cat ~/.claude/settings.local.json 2>/dev/null | head -5
```
If hooks aren't firing, check:
- hooks.json is in the right location
- Script paths use `${CLAUDE_PLUGIN_ROOT}` or absolute paths
- Scripts have execute permissions
### 2a. Deterministic Hook Sanity
Verify each hook script on both happy and failing inputs:
```bash
# commit-validate: valid passes, bad format blocks (exit 2)
echo '{"tool_input":{"command":"git commit -m \"feat: x\""}}' | node "$CLAUDE_PLUGIN_ROOT/scripts/commit-validate.js" && echo "commit-validate pass: OK"
echo '{"tool_input":{"command":"git commit -m \"not conventional\""}}' | node "$CLAUDE_PLUGIN_ROOT/scripts/commit-validate.js" 2>&1; test $? -eq 2 && echo "commit-validate block: OK"
# secret-scan: clean passes, hardcoded key blocks
echo '{"tool_input":{"content":"hello"}}' | node "$CLAUDE_PLUGIN_ROOT/scripts/secret-scan.js" && echo "secret-scan pass: OK"
echo '{"tool_input":{"content":"key=\"AKIAIOSFODNN7EXAMPLE\""}}' | node "$CLAUDE_PLUGIN_ROOT/scripts/secret-scan.js" 2>&1; test $? -eq 2 && echo "secret-scan block: OK"
# git-blast-radius: safe passes, destructive blocks
echo '{"tool_input":{"command":"git status"}}' | node "$CLAUDE_PLUGIN_ROOT/scripts/git-blast-radius.js" && echo "git-blast-radius pass: OK"
echo '{"tool_input":{"command":"git reset --hard"}}' | node "$CLAUDE_PLUGIN_ROOT/scripts/git-blast-radius.js" 2>&1; test $? -eq 2 && echo "git-blast-radius block: OK"
```
Every line should print an `OK`. Missing any means the script is silently broken.
### 2b. Git Safety Override
If you need to run a blocked git operation deliberately (e.g. intentional
`git reset --hard` during recovery), export `PRO_WORKFLOW_ALLOW_UNSAFE_GIT=1`
for the shell that will run it. Never set it globally.
### 3. Context Health
```text
/context
```
- Usage < 70%: Healthy
- Usage 70-90%: Consider `/compact`
- Usage > 90%: Compact immediately or start fresh
### 4. MCP Servers
Check active MCPs. Target: <10 servers, <80 tools.
### 5. CLAUDE.md Size
```bash
wc -l CLAUDE.md 2>/dev/null || echo "No CLAUDE.md found"
```
- < 60 lines: Ideal
- 60-150 lines: Acceptable
- > 150 lines: Too long, split into modules
### 6. Git Status
```bash
git status --short
git stash list
```
- Uncommitted changes? Commit or stash.
- Stale stashes? Clean up.
### 7. Settings Validation
```bash
cat .claude/settings.json 2>/dev/null | head -3
cat ~/.claude/settings.json 2>/dev/null | head -3
```
- Check for conflicting settings between project and user level
- Verify permission rules are correct
### 8. Wiki Knowledge Base
```bash
node $PRO_WORKFLOW_ROOT/skills/wiki-builder/scripts/wiki-cli.js list 2>/dev/null
node $PRO_WORKFLOW_ROOT/skills/wiki-research-loop/scripts/research-loop.js status 2>/dev/null
ls -1 ~/.pro-workflow/STOP 2>/dev/null && echo "KILL SWITCH ACTIVE — no research runs"
tail -5 ~/.pro-workflow/tick.log 2>/dev/null
```
- List of wikis · seed counts per status · kill-switch state · last cron-tick activity
- Embeddings require an OpenAI or Voyage key configured for this plugin. BM25 search remains local and needs no key. Never print environment values or read credential files.
### 9. Council Providers
In a plugin session, call `run_provider_task` on the `providers` MCP server with `{"task":"council","args":["providers"]}`. It reports configured status without revealing values. The command below is for standalone CLI use.
```bash
node $PRO_WORKFLOW_ROOT/skills/llm-council/scripts/council.js providers 2>/dev/null
```
- Shows which pro-workflow credentials are configured (Anthropic/OpenAI/OpenRouter/Fireworks/custom), without exposing values.
## Quick Fixes
| Issue | Fix |
|-------|-----|
| Commands missing | `cp -r /path/to/pro-workflow/commands/* ~/.claude/commands/` |
| Database missing | `mkdir -p ~/.pro-workflow && npm run build` (in plugin dir) |
| Hooks not firing | Check paths in hooks.json use absolute paths |
| Context degraded | Run `/compact` or start fresh session |
| Too many MCPs | Disable unused: `disabledMcpjsonServers` in settings |
| CLAUDE.md too long | Split into root + package-level files |
## Report
After running checks, summarize:
```text
Pro-Workflow Health Check
Installation: OK / NEEDS SETUP
Hooks: ACTIVE / INACTIVE
Context: XX% (healthy/warning/critical)
MCPs: X active (OK / TOO MANY)
CLAUDE.md: XX lines (OK / SPLIT RECOMMENDED)
Git: clean / X uncommitted files
```
© Rohit Ghumare · MIT · rohitg00/pro-workflow
Similar slash commands
All slash commands for Agent setup: memory, context, orchestration →doctor by other authors comes first.