| name | debug-desktop |
| description | Debug the desktop Electron app — tail logs, check log status, diagnose errors, review recent renderer/agent/electron logs. |
| argument-hint | [tail | status | errors | renderer | agent | electron | all] |
| allowed-tools | ["Bash","Read","Grep","Glob"] |
Desktop App Debugger
You are a debugging assistant for the Claude for Beginners desktop app (Electron + Next.js).
Log Directory
All logs live at ~/.claude-for-beginners/logs/:
| File | Source | What's in it |
|---|
electron.log | Electron main process | App lifecycle, window events, IPC calls, Next.js server stdout/stderr |
agent.log | Next.js /api/agent route | Claude CLI spawn, streaming messages, session lifecycle |
app.log | Next.js server-side code | General application logs |
renderer.log | Browser renderer (via IPC) | console.error, console.warn, uncaught exceptions, unhandled promise rejections |
Rotated files: electron.log.old, agent.log.1 through .5, etc.
Commands by Argument
Parse $ARGUMENTS and run the matching action:
tail (or no argument)
Stream all logs live. Run:
tail -f ~/.claude-for-beginners/logs/*.log
Use run_in_background: true so the user can keep chatting. Check output periodically with TaskOutput.
status
Show all log files with sizes and last-modified times:
ls -lhtr ~/.claude-for-beginners/logs/ 2>/dev/null || echo "No log directory found"
Summarize what you find: which logs exist, which are recent, which are large.
errors
Search ALL log files for recent errors:
grep -n "ERROR\|error\|Error\|WARN\|warn\|Uncaught\|unhandled\|ENOENT\|EACCES\|FATAL\|panic\|crash" ~/.claude-for-beginners/logs/*.log | tail -50
Group and summarize the errors found. Highlight the most recent and most frequent ones.
renderer
Show the last 80 lines of renderer.log:
tail -80 ~/.claude-for-beginners/logs/renderer.log 2>/dev/null || echo "No renderer.log — the app may not have been started yet, or no errors/warnings have occurred."
Analyze what you see: are there React errors? Unhandled rejections? Repeated warnings?
agent
Show the last 80 lines of agent.log:
tail -80 ~/.claude-for-beginners/logs/agent.log 2>/dev/null || echo "No agent.log — no agent sessions have run yet."
Look for: Claude CLI spawn failures, streaming errors, unexpected session ends.
electron
Show the last 80 lines of electron.log:
tail -80 ~/.claude-for-beginners/logs/electron.log 2>/dev/null || echo "No electron.log — the app hasn't been started yet."
Look for: startup failures, window load errors, IPC issues, Next.js server problems.
all
Read the last 40 lines of every log file and provide a unified summary:
for f in ~/.claude-for-beginners/logs/*.log; do echo "=== $(basename $f) ==="; tail -40 "$f" 2>/dev/null; echo; done
Cross-reference events across files. Build a timeline if possible.
Free-form text (anything else)
Treat $ARGUMENTS as a search query. Grep all log files for it:
grep -rn "$ARGUMENTS" ~/.claude-for-beginners/logs/ | tail -50
Analysis Guidelines
When reviewing logs:
- Timestamps first — establish when things happened and in what order
- Cross-reference — an error in
renderer.log often has a corresponding entry in agent.log or electron.log
- Patterns — repeated errors matter more than one-offs
- Recency — focus on the most recent entries unless the user asks about something specific
- Actionable output — always end with what to do next (fix suggestion, file to check, or "looks healthy")
Source Code References
If you need to trace an error back to source code:
- Electron main process:
desktop/electron/main.js
- Preload bridge:
desktop/electron/preload.js
- Renderer logger:
desktop/src/lib/renderer-logger.ts
- Server-side logger:
desktop/src/lib/logger.ts
- Agent API route:
desktop/src/app/api/agent/route.ts
