name: completion-check
description: "Completion Check: Verify Infrastructure Is Wired"
user-invocable: false

Completion Check: Verify Infrastructure Is Wired

When building infrastructure, verify it's actually connected to the system before marking as complete.

Pattern

Infrastructure is not done when the code is written - it's done when it's wired into the system and actively used. Dead code (built but never called) is wasted effort.

DO

1. Trace the execution path - Follow from user intent to actual code execution:
   bash # Example: Verify Task tool spawns correctly grep -r "claude -p" src/ grep -r "Task(" src/

2. Check hooks are registered, not just implemented:
   ```bash
   # Hook exists?
   ls -la .claude/hooks/my-hook.sh

# Hook registered in settings?
grep "my-hook" .claude/settings.json
```

1. Verify database connections - Ensure infrastructure uses the right backend:
   bash # Check connection strings grep -r "postgresql://" src/ grep -r "sqlite:" src/ # Should NOT find if PostgreSQL expected

2. Test end-to-end - Run the feature and verify infrastructure is invoked:
   ```bash
   # Add debug logging
   echo "DEBUG: DAG spawn invoked" >> /tmp/debug.log

# Trigger feature
uv run python -m my_feature

# Verify infrastructure was called
cat /tmp/debug.log
```

1. Search for orphaned implementations:
   bash # Find functions defined but never called ast-grep --pattern 'async function $NAME() { $$$ }' | \ xargs -I {} grep -r "{}" src/

DON'T

• Mark infrastructure "complete" without testing execution path
• Assume code is wired just because it exists
• Build parallel systems (Task tool vs claude -p spawn)
• Use wrong backends (SQLite when PostgreSQL is architected)
• Skip end-to-end testing ("it compiles" ≠ "it runs")

Completion Checklist

Before declaring infrastructure complete:

• [ ] Traced execution path from entry point to infrastructure
• [ ] Verified hooks are registered in .claude/settings.json
• [ ] Confirmed correct database/backend in use
• [ ] Ran end-to-end test showing infrastructure invoked
• [ ] Searched for dead code or parallel implementations
• [ ] Checked configuration files match implementation

Example: DAG Task Graph

Wrong approach:

✓ Built BeadsTaskGraph class
✓ Implemented DAG dependencies
✓ Added spawn logic
✗ Never wired - Task tool still runs instead
✗ Used SQLite instead of PostgreSQL

Right approach:

✓ Built BeadsTaskGraph class
✓ Wired into Task tool execution path
✓ Verified claude -p spawn is called
✓ Confirmed PostgreSQL backend in use
✓ Tested: user calls Task() → DAG spawns → beads execute
✓ No parallel implementations found

Source Sessions

• This session: Architecture gap discovery - DAG built but not wired, Task tool runs instead of spawn, SQLite used instead of PostgreSQL