redsub-validate

name	redsub-validate
description	Run lint, type check, and unit tests with evidence.

Code Validation

Command Resolution

Determine the project's validation commands:

1. Check project CLAUDE.md for explicit commands (lint, check, test)
2. If not found, read package.json scripts and infer:
   • Lint: script containing "lint" (e.g., lint, lint:fix)
   • Type check: script containing "check" (e.g., check, typecheck)
   • Unit test: script containing "test" (e.g., test, test:unit)
3. Detect package manager: look for lock files (pnpm-lock.yaml → pnpm, yarn.lock → yarn, default → npm)
4. If ambiguous, ask the user

Procedure

Run sequentially. Stop on first failure.

1. Lint

<package-manager> run <lint-script>

2. Type check

<package-manager> run <check-script>

3. Unit tests

<package-manager> run <test-script>

If the test script doesn't include a --run flag for watch-mode frameworks (e.g., vitest), append -- --run.

4. SSOT Consistency Checks

When a canonical source exists (config, constants, types), verify consumers stay in sync:

• Config-driven values: verify runtime reads match the canonical source (e.g., routes file exports match router config)
• Shared types: verify API response shapes match the declared type (snapshot or schema validation)
• i18n: verify all keys used in components exist in translation files
• If a test duplicates a magic number or string, extract to shared fixture or import from source module

5. Evidence (5-Step Verification Gate)

No claims without evidence. Follow this exact sequence:

1. IDENTIFY: What command proves this claim?
2. RUN: Execute the full command (fresh, complete)
3. READ: Check entire output, exit code, failure count
4. VERIFY: Does the output support the claim?
5. CLAIM: Only then make the claim

Forbidden before verification: "should", "probably", "seems to", or satisfaction expressions ("Great!", "Done!", "Perfect!").

On success:

Validation passed (with evidence):
- lint: pass (0 errors, 0 warnings)
- type check: pass (0 errors)
- unit test: pass (N tests, 0 failures)

On failure:

Validation failed:
- [step]: [error details]
- Fix: [suggestion]