The Virtual Board of Directors for AI-First Developers and Vibe Coders.
A production-grade, open-source library of 13 specialized role-based audit prompts designed to turn 10x AI prototypes into battle-hardened, production-ready software.
"Vibe coding" (a phrase popularized by Andrej Karpathy) describes the shift where developers build applications primarily by prompting AI coding models (Cursor, Antigravity, Claude Code, Windsurf, Copilot, ChatGPT, etc.) while acting as high-level executive directors.
While AI agents let you ship prototypes at 10x speed, most vibe-coded apps hit the Prototype Wall when moving to production:
- Security Flaws: Missing CORS rules, unvalidated webhooks, raw SQL concatenation, IDOR bugs.
- UX Churn: Confusing onboarding, missing empty states, bad error messages that drive users away.
- DevOps Hazards: Scale-to-zero misconfigurations, missing connection pooling, non-stateless containers.
- AI Vulnerabilities: Unprotected prompts, unbudgeted token consumption, raw JSON parsing crashes.
The Vibe Coder Prompt Library solves this by providing 13 domain-expert roles—from CISO Red Teaming to Principal PM and Database Resilience—that rigorously audit your codebase before you launch.
| # | Role / Persona Prompt | Focus & Evaluation Standard | Best Used On |
|---|---|---|---|
| 01 | 01_principal_product_manager.md |
15 PLG & Onboarding Pillars: Time-to-value (TTV), habit loops, feature bloat elimination. | Journey Maps, Onboarding Flows |
| 02 | 02_head_of_growth_copywriting.md |
15 CRO Copy Criteria: H1 hooks, objection preemption, risk reversal, CTA optimization. | Landing & Pricing Pages, Ads |
| 03 | 03_ciso_red_team.md |
35-Point Red Team Pentest: SQLi, IDOR, JWT confusion, webhook HMAC, SSRF, prompt injection. | Backend APIs, Auth, Gateways |
| 04 | 04_head_of_customer_success.md |
12-Point Frictionless CX Standard: Support-prevention diagnostics, plain English errors, empty states. | Error Alerts, Empty States |
| 05 | 05_devops_infra_sre.md |
15 Lean SRE Criteria: Zero-downtime deploys, connection pooling, resource tuning, secrets. | Dockerfiles, Terraform, CI/CD |
| 06 | 06_ui_ux_polish.md |
15 UI/UX Polish Standards: Typography hierarchy, CLS prevention, accessibility, touch targets. | Frontend Components, CSS/Tailwind |
| 07 | 07_production_readiness_adversarial.md |
20 Production Pillars: Adversarial red-team checks, cookie flags, rate limits, health probes. | Full Codebase Pre-Deployment |
| 08 | 08_compliance_privacy_auditor.md |
15 Regulatory Criteria: GDPR, CCPA, HIPAA, COPPA, SOC 2, PCI-DSS compliance verification. | DB Schemas, Privacy Flows |
| 09 | 09_eval_framework_architect.md |
15 LLM Reliability Pillars: RAG relevance, hallucination guardrails, Promptfoo e2e evals, caching. | AI Workflows, System Prompts |
| 10 | 10_database_resilience_disaster_recovery.md |
15 Database Recovery Criteria: Point-in-time recovery, non-blocking migrations, PgBouncer. | SQL Schemas, DB Configs |
| 11 | 11_ci_cd_secops_pipeline.md |
15 DevSecOps Quality Gates: Commit secret scans, SAST/DAST, OIDC authentication, frozen locks. | GitHub Actions, CI/CD YAML |
| 12 | 12_sr_principal_software_engineer.md |
15 Engineering Quality Pillars: DRY architecture, 70%+ test coverage, JSON logging, OpenTelemetry. | Architecture & Code Quality |
| 13 | 13_patent_attorney.md |
Software Patent Specification: Drafts detailed invention descriptions under Alice framework. | Architecture & Tech Notes |
- Open your target file or directory in your AI coding tool.
- Copy the contents of any prompt from the
prompts/directory. - Paste it as a system prompt or user message, replacing
[INSERT APP NAME]with your project's name. - Let the AI model execute the multi-pillar audit and generate exact code/config fixes!
You can run automated code audits locally against your codebase using Ollama and open-weights models (e.g. qwen2.5-coder, llama3.3, deepseek-r1):
# 1. Clone the repository
git clone https://github.com/samirsss/vibe-coder-prompt-library.git
cd vibe-coder-prompt-library
# 2. Run an audit on a local project directory
python3 scripts/run_review.py \
--prompt prompts/03_ciso_red_team.md \
--target ../my-awesome-app/src \
--app "MyAwesomeApp" \
--model "qwen2.5-coder:32b" \
--output audit_report.mdContributions are welcome! If you have optimized role prompts, new audit criteria, or runner integrations, feel free to open a Pull Request.
This repository is licensed under the MIT License.