Construct-AI-primary Validator - Syntax & Logic Validation
Overview
Validator validates prompt syntax and logic for PromptForge AI. A single syntax error, unclosed bracket, or logical contradiction can cause a prompt to fail silently or produce garbage output. Validator catches these issues before prompts reach production.
When to Use
- Before deploying any new prompt to production
- After modifying an existing prompt (even small changes)
- When troubleshooting unexpected prompt behavior
- When reviewing prompts created by other agents
- When building automated prompt validation pipelines
- Don't use when: Testing prompt output quality (use quantifier-promptforge-benchmarking)
Core Procedures
Step 1: Syntax Validation
Check the prompt's structural integrity:
- All brackets, braces, and quotes are properly closed
- Variable placeholders use consistent syntax (
{{variable}}or{variable}) - No orphaned or dangling references (references to undefined variables)
- Markdown/formatting syntax is valid
- No invisible characters or encoding issues
- Prompt length is within model limits
Step 2: Logical Consistency Check
Verify the prompt makes logical sense:
- Instructions don't contradict each other
- Output format specification is achievable
- Constraints are mutually compatible (e.g., not "be brief" AND "include all details")
- Role definition is consistent with instructions
- Examples match the described behavior
- Edge cases are handled (what if input is empty? too long?)
Step 3: Completeness Check
Ensure nothing critical is missing:
- System role is defined
- Task/instructions are present
- Input format is specified
- Output format is specified
- Constraints and guardrails are included
- Error handling instructions (what to do if input is invalid)
Step 4: Model Compatibility Check
Verify the prompt will work with target models:
- Prompt length fits within model's context window
- No model-specific features used incorrectly
- Temperature and other parameters are valid for the model
- Special tokens or formatting are supported by the model
Step 5: Validation Report
VALIDATION REPORT
=================
Prompt: [name and version]
Validator: [agent name]
Date: [validation date]
Syntax: [PASS/FAIL - issues found]
Logic: [PASS/FAIL - issues found]
Completeness: [PASS/FAIL - issues found]
Model Compatibility: [PASS/FAIL - issues found]
Overall: [PASS/FAIL]
Issues: [list of all issues with severity]
- [CRITICAL] [issue description]
- [WARNING] [issue description]
- [INFO] [issue description]
Recommendation: [deploy/fix before deploy]
Success Metrics
- Validation coverage: 100% of production prompts pass validation
- Critical issue detection: 100% of syntax errors caught before deployment
- False positive rate: <5% (valid prompts incorrectly flagged)
- Validation time: <5 minutes per prompt
Error Handling
- Error: Critical syntax issue found Response: Do NOT deploy, fix the syntax issue, re-validate
- Error: Logical contradiction found Response: Do NOT deploy, resolve the contradiction (clarify which instruction takes priority), re-validate
- Error: Prompt passes validation but still produces poor output Response: Validation catches syntax/logic issues, not quality issues—transfer to clarity-promptforge-performance-optimization for quality improvement
Cross-Team Integration
Related Skills: blueprint-promptforge-template-designer, refiner-promptforge-version-control, integration-promptforge-cross-agent-compatibility, output-validation-checklist Used By: Blueprint, Refiner, Integration, ALL agents creating or modifying prompts