poteto Oops
Fix infrastructure failures. Applies to both user-project code and Noodle-internal state.
Fix flow
Reproduce -> Diagnose -> Fix -> Verify -> Commit
1. Reproduce
Run the failing command. Capture exact error output. If intermittent, run multiple times to confirm.
2. Diagnose
Suspect state before code. Check in this order:
- Noodle state —
.noodle/files, queue consistency, config validity - Environment — Go version, missing deps, stale caches, tmux session health
- Config —
.noodle.toml(project root),.agents/skills/structure,brain/integrity - Persistent files — lock files, temp files, incomplete writes
- Code — only after ruling out state and environment
Principles
- [[fix-root-causes]]
- [[prove-it-works]]
Noodle-internal checklist:
.noodle/state files: corrupt, stale, or inconsistent?- Queue state: orphaned tasks, stuck sessions?
- Config: valid TOML, referenced paths exist?
- tmux sessions: zombie processes, detached sessions?
3. Fix
Trace to root cause. Never paper over symptoms.
After fixing, grep for the same pattern elsewhere — the bug may exist in multiple locations.
4. Verify
Run the original failing command. Confirm it passes. Run related tests to check for regressions.
5. Commit
Commit message describes the root cause, not symptoms.
# Good: fix(queue): stale lock file prevented task dequeue after crash
# Bad: fix: tasks not running