/quality-checks

description: "Run all quality checks (build, test, lint, typecheck, Hono inference) before completing tasks"
allowed-tools: ["bash"]

Quality Checks

Runs all pre-commit validation checks to ensure code quality before task completion.

What Gets Checked

1. **Build** - Verify compilation succeeds 2. **Tests** - Ensure all tests pass (unit + integration) 3. **Lint** - Check code style and conventions 4. **Typecheck** - Validate TypeScript types 5. **Hono Inference** - Ensure exported typed Hono clients still infer concrete JSON payloads

Usage

/quality-checks

Implementation

Use the `scripts/quality-checks.sh` script to run individual checks or all checks.

Script Usage

# Run individual checks (default: show last 200 lines on failure)
bash scripts/quality-checks.sh build
bash scripts/quality-checks.sh test
bash scripts/quality-checks.sh lint
bash scripts/quality-checks.sh typecheck
bash scripts/quality-checks.sh hono-inference

# Control output on failure (optional second argument)
bash scripts/quality-checks.sh build 100      # Show last 100 lines
bash scripts/quality-checks.sh test 50        # Show last 50 lines
bash scripts/quality-checks.sh typecheck all  # Show all output

# Run all checks in order
bash scripts/quality-checks.sh all            # Show last 200 lines on failure
bash scripts/quality-checks.sh all 100        # Show last 100 lines on failure
bash scripts/quality-checks.sh all all        # Show all output on failure

Instructions for LLM

**Run checks in this order and stop on first failure:**

1. Build check: `bash scripts/quality-checks.sh build` 2. Test check: `bash scripts/quality-checks.sh test` 3. Lint check: `bash scripts/quality-checks.sh lint` 4. Typecheck: `bash scripts/quality-checks.sh typecheck` 5. Hono inference: `bash scripts/quality-checks.sh hono-inference`

**Default behavior:**

• Shows last 200 lines on failure (usually sufficient)
• No output on success (saves LLM context)
• You can request more lines if needed: `bash scripts/quality-checks.sh build 500`
• You can request all output if needed: `bash scripts/quality-checks.sh build all`

**Flexibility:**

• Run individual checks as needed (e.g., just build after code changes)
• Re-run specific failing checks
• Adjust output lines based on error complexity
• Use `bash scripts/quality-checks.sh all` for convenience to run all checks

**Output Strategy:**

• Each command runs **once only** (no double-runs)
• Output captured to `/tmp/build/dexto-{check}-$$.log`
• **Success**: No output shown, temp file deleted (minimal context)
• **Failure**: Last N lines displayed (default 200), temp file deleted, execution stops

This minimizes unnecessary context while providing sufficient diagnostics on failures.

Output Format

**On success** (minimal output):

Running quality checks...

✅ Build passed
✅ Tests passed
✅ Lint passed
✅ Typecheck passed
✅ Hono Inference passed

All quality checks passed! ✨

**On failure** (full diagnostic output):

Running quality checks...

✅ Build passed
❌ Tests failed:

[Full test output displayed here]

When to Run

• **Before completing any task** - Mandatory
• **Before creating commits** - Ensures clean history
• **After major refactoring** - Verify nothing broke
• **Before requesting PR review** - Save reviewer time
• **After changing any Hono route/client contract** - Catch inference regressions before downstream repos do

Notes

• All checks must pass before marking tasks as complete
• The command exits on first failure for fast feedback
• Lint warnings don't block completion, but should be addressed
• Tests include both unit and integration tests