Now liveThe Skillselion MCP - thousands of ranked skills, loaded into your agent mid-task. No install.Get it →
ductrung-nguyen avatar

External Validation

  • 1 installs
  • Updated February 22, 2026
  • ductrung-nguyen/orca-vscode

Invoke a different LLM to give unbiased critique of PRDs and implementation plans, then merge internal and external feedback with fallback handling.

About

Defines a protocol for calling an external LLM of a different model family to critique PRDs and plans, then merging feedback. A developer uses it to reduce model blindness when validating artifacts in a spec pipeline.

  • Model-selection rule requires a different validator family
  • Feedback merging fails if either validator fails

External Validation by the numbers

  • 1 all-time installs (skills.sh)
  • Ranked #14,102 of 16,546 AI & Agent Building skills by installs in the Skillselion catalog
  • Data as of Jul 8, 2026 (Skillselion catalog sync)
npx skills add https://github.com/ductrung-nguyen/orca-vscode --skill external-validation

Add your badge

Show developers this skill is listed on Skillselion. Paste this into your README.

Listed on Skillselion
Installs1
Last updatedFebruary 22, 2026
Repositoryductrung-nguyen/orca-vscode

What it does

Invoke a different LLM to give unbiased critique of PRDs and implementation plans, then merge internal and external feedback with fallback handling.

Files

SKILL.mdMarkdownGitHub ↗

External Validation

Protocol for invoking external LLM validation to provide unbiased critique of artifacts.

Core Principle

WHY External Validation: Internal validation (same model that created the artifact) can suffer from:

  • Model blindness (model won't catch its own mistakes)
  • Confirmation bias (validating own reasoning)
  • Pattern repetition (same model, same blind spots)

External validation uses a DIFFERENT model to provide fresh perspective.

When to Use External Validation

StageArtifactRequired?
validate-prdPRDOptional but recommended
validate-planImplementation PlanOptional but recommended
validate-implImplementationRequired (code tools)

Model Selection Strategy

Rule: External validator MUST be a different model family than the creator.

Creator ModelRecommended Validator
GPT-5.2Claude Opus or Gemini
Claude OpusGPT-5.2 or Gemini
Claude SonnetGPT-5.2 or Gemini
GeminiGPT-5.2 or Claude Opus

Feedback Merging Protocol

When combining internal and external validation feedback:

1. Status: FAIL if either validator fails 2. Issues: Combine and deduplicate from both validators 3. Trust external: If external finds critical issues internal missed

Status File Integration

{
  "stages": {
    "validate-prd": {
      "status": "completed",
      "output": "PRD validated",
      "validator": "internal + gemini-3.0-pro"
    }
  },
  "externalValidation": {
    "prdValidation": {
      "status": "completed",
      "model": "gemini-3.0-pro",
      "result": "PASS",
      "issues": 0
    }
  }
}

Error Handling

If external service is unavailable:

  • Fall back to internal validation only
  • Record skipped status with reason
  • Continue pipeline with warning

Anti-Patterns

  • ❌ Using same model for external validation
  • ❌ Skipping external validation on complex PRDs
  • ❌ Ignoring external feedback when internal passes

Environment Variables

EXTERNAL_VALIDATE=true              # Enable external validation
EXTERNAL_VALIDATION_MODEL="gemini-3.0-pro"  # Model to use
EXTERNAL_VALIDATION_TIMEOUT=30000   # 30 seconds

Related skills

This week in AI coding

Five minutes, every Monday - the tools, releases and tactics for developers.

unsubscribe anytime.