quality-gates-runner

Files (1)

• SKILL.md

SKILL.md (first lines)

# Quality Gates Runner

Run quality checks via subagents to keep main context clean.

## When to Use

- After implementation is complete
- Before creating a PR
- In target repository directory

## Prerequisites

1. In target repository directory (auto-checkout below if PR URL/number given)
2. Changes exist (unstaged or staged)
3. `gh` CLI authenticated (if using auto-checkout)

## Workflow

### 0. Checkout PR Branch (if given PR URL/number)

If not already in the target repo, check out the branch:

```bash
# Extract PR info
PR_NUMBER=<N>
REPO_NAME=$(gh pr view "$PR_NUMBER" --json repository --jq '.repository.name // empty' 2>/dev/null \
  || basename "$(gh repo view --json name -q .name)")
BRANCH=$(gh pr view "$PR_NUMBER" --json headRefName -q .headRefName)
```

**Try methods in order:**

```bash
# Method 1: Already on the branch
CURRENT=$(git branch --show-current 2>/dev/null)
if [ "$CURRENT" = "$BRANCH" ]; then
  echo "Already on $BRANCH"

# Method 2: Existing worktree
elif [ -d "${WORKTREE_BASE:-$HOME/worktrees}/$REPO_NAME/$BRANCH" ]; then
  cd "${WORKTREE_BASE:-$HOME/worktrees}/$REPO_NAME/$BRANCH"

# Method 3: Cross-repo task directory
elif ls -d "${CROSS_REPO_BASE:-$HOME/cross-repo-tasks}"/*/"$REPO_NAME" 2>/dev/null | head -1 | grep -q .; then
  cd "$(ls -d "${CROSS_REPO_BASE:-$HOME/cross-repo-tasks}"/*/"$REPO_NAME" 2>/dev/null | head -1)"

# Method 4: wt-continue (branch exists on remote)
else
  wt-continue "$REPO_NAME" "$BRANCH"
  cd "${WORKTREE_BASE:-$HOME/worktrees}/$REPO_NAME/$BRANCH"
fi
```

### 1. Verify Directory

Confirm in target repository and check for changes:

```bash
pwd  # Should be in target repository
git status --short  # Should show modified files
```

### 2. Dispatch Subagents in Parallel

Create tasks for each quality gate using the Task tool. Dispatch all in parallel:

**Task 1: Lint**

```
Run the lint command (e.g., `pnpm lint` or per AGENTS.md) in the target repository.
Report: PASS or FAIL with specific errors and file:line locations.
If FAIL, suggest specific fixes for each error.
```

**Task 2: Format**

```
Run the format check command (e.g., `pnpm format:check`) in the target repository.
Report: PASS or FAIL.
If FAIL, run the format fix command and report files changed.
```

**Task 3: Typecheck**

```
Run the typecheck command (e.g., `pnpm typecheck`) in the target repository.
Report: PASS or FAIL with specific type errors and locations.
For each error, suggest the fix.
```

**Task 4: Unused Code Check** (if available)

```
Run unused code detection (e.g., `pnpm knip`) if configured.
Report: PASS or FAIL with unused exports/dependencies found.
Suggest removals for each finding.
```

**Task 5: Unit Tests**

```
Run unit tests (e.g., `pnpm test:unit` or per AGENTS.md) in the target repository.
Report: PASS with test count, or FAIL with failing test names and errors.
For failures, include the assertion that failed.
```

**Task 6: Style Lint** (only if CSS changes exist)

```
Run style linting (e.g., `pnpm stylelint`) if configured.
Report: PASS or FAIL with issues found.
```

### 3. Collect and Report Results

Wait for all subagents, then compile report:

```markdown
# Quality Gates Report

From DrJKL/ticket-to-pr-pipeline, MIT.

More from DrJKL/ticket-to-pr-pipeline