Parallel Validator

Run multiple validation tasks (tests, lint, security) in parallel background workers to minimize wait time and keep main context clean.

When To Use

Auto-Trigger Signals:

• Code changes complete and ready for validation
• Multiple validators applicable (tests + lint + security)
• Before commit/PR creation
• User says "validate", "check everything", "run all checks"

Skip When:

• Only one validator needed (use that skill directly)
• Quick single-file change (overhead not worth it)
• User needs real-time feedback on specific test

Inputs

• Code changes to validate
• Which validators to run (auto-detected or specified)
• Pass/fail thresholds

Outputs

• Aggregated pass/fail status
• Summary of issues by severity
• Detailed results available on request

Validator Detection

Auto-detect applicable validators by checking for config files:

Parallel Execution Pattern

Step 1: Detect Validators

# Check what's available
ls pytest.ini pyproject.toml 2>/dev/null && echo "pytest"
ls jest.config.* vitest.config.* 2>/dev/null && echo "jest"
ls .eslintrc* 2>/dev/null && echo "eslint"
ls tsconfig.json 2>/dev/null && echo "typescript"

Step 2: Spawn All in Parallel

In a SINGLE message, spawn all validators as background tasks:

# Spawn all in one message (parallel execution)
Task:
  subagent_type: background-worker
  description: "Run pytest"
  prompt: "Run pytest with -v flag. Return pass/fail count and any failures."
  run_in_background: true

Task:
  subagent_type: background-worker
  description: "Run ESLint"
  prompt: "Run eslint on src/. Return error count and top 5 issues."
  run_in_background: true

Task:
  subagent_type: security-auditor
  description: "Security scan"
  prompt: "Scan for OWASP top 10, secrets, auth issues. Return severity-ranked findings."
  run_in_background: true

Step 3: Continue Work (Optional)

While validators run, you can continue with other tasks if applicable.

Step 4: Poll for Results

# Check all validators (can poll in parallel too)
TaskOutput: { task_id: "pytest_id", block: true }
TaskOutput: { task_id: "eslint_id", block: true }
TaskOutput: { task_id: "security_id", block: true }

Step 5: Aggregate and Report

## Validation Results

| Validator | Status | Summary |
|-----------|--------|---------|
| pytest | PASS | 42 tests passed |
| eslint | WARN | 3 warnings (no errors) |
| security | PASS | No vulnerabilities found |

**Overall: READY FOR COMMIT**

Failure Handling

Critical Failures (Block Commit)

• Test failures
• Security vulnerabilities (HIGH/CRITICAL)
• Type errors

Warnings (Allow Commit with Note)

• Lint warnings
• Low-severity security findings
• Coverage below threshold (but not failing)

Action on Failure

1. Report which validators failed
2. Show specific failures (not full output)
3. Ask: "Fix issues or commit anyway?"
4. If fix: Address failures, re-run only failed validators
5. If commit anyway: Document skipped checks in commit message

Context Efficiency

Token Savings

Each validator agent:

• Runs in isolated context
• Only returns summary (not full logs)
• ~200 tokens per validator returned to main

vs Traditional:

• Full pytest output (5000+ tokens)
• Full eslint output (2000+ tokens)
• Full security report (3000+ tokens)

Savings: 90%+ context reduction

Integration with Other Skills

With implement-plan

After each task group completion, trigger parallel-validator:

Task group complete → parallel-validator → report → next group

With git-workflow

Before commit/PR, auto-run:

git add → parallel-validator → if pass → commit

With beads

Track validator runs as agent tasks:

bd create "Validation run" -t agent_task --meta '{"validators": ["pytest", "eslint", "security"]}'

Command Shortcuts

# Validate everything
/parallel-validator

# Validate specific
/parallel-validator tests,lint

# Validate with coverage
/parallel-validator --with-coverage

# Quick validation (skip security)
/parallel-validator --quick

Anti-Patterns

• Running validators sequentially when parallel is possible
• Waiting for each validator before starting next
• Including full test output in main context
• Re-running all validators when only one failed
• Skipping security check to save time

Keywords

parallel, validate, tests, lint, security, background, concurrent, pre-commit, ci, check