SkillProof AI

💡 Inspiration

Generative AI has made producing answers virtually zero-cost. Students can instantly generate essays, code, and project plans. However, this creates a massive problem for educational institutions, hackathon judges, and employers: how do we verify actual capability? We were inspired to shift the focus from generating answers to generating proof. We wanted to build an agentic lab that acts as a strict, evidence-seeking mentor rather than just another chatbot that does the homework for you.

🚀 What it does

SkillProof AI turns any learning goal into a structured, verifiable project mission. It guides the learner through the process, actively requesting concrete evidence (like repository commits, architecture diagrams, and reflection notes).

Key features include:

• Mission Architect & Mentor Agents: Break down goals into step-by-step tasks.
• Proof Lab: A centralized hub to collect, structure, and evaluate evidence.
• Strict Jury Review: A local evaluation API that scores the submitted evidence against official criteria (e.g., Innovation, Technical Complexity, Real-world Impact).
• Portfolio Packager: Automatically exports a Devpost-ready proof pack containing all validated artifacts.

🛠️ How we built it

We built the platform using Next.js 16, React 19, and TypeScript, creating a fast, zero-config deployable web application. For styling and UI, we used modern CSS with Lucide React icons.

The core intelligence is powered by a deterministic multi-agent strategy kernel. We designed an evidence graph that maps required proof signals to the official FutureAI hackathon criteria. We also integrated an optional OpenAI-compatible API adapter (supporting models like GPT-4) to allow for dynamic, AI-assisted validation when API keys are present, while ensuring the MVP works perfectly without external credentials using our deterministic fallback.

🚧 Challenges we ran into

The biggest challenge was orchestrating specialist agents with strict guardrails. It's easy to make an AI generate a lot of text, but it is technically complex to force the AI to evaluate external evidence objectively without hallucinating success. We had to build a strict rule gate and an artifact audit script (qa:artifact-audit) to keep the agents disciplined and ensure they separate "local merit" from actual "public readiness".

🏆 Accomplishments that we're proud of

We are incredibly proud of the Strict Jury Stress Test API. By implementing a system that actively searches for missing evidence and flags weak claims before submission, we built a tool that genuinely improves the quality and integrity of a student's portfolio. We are also proud that the entire app is optimized for zero-config Vercel deployment.

📚 What we learned

We learned that prompting for "evaluation" requires entirely different architecture than prompting for "generation". We had to deeply understand the official 30/25/20/15/10 rubric of the FutureAI hackathon and translate those human-judging criteria into machine-readable JSON contracts and agent validation rules.

⏭️ What's next for SkillProof AI

Our next major milestones include:

1. GitHub API Integration: Automatically pulling commit history and PRs as verifiable evidence.
2. LMS Integration: Connecting with platforms like Canvas or Moodle for institutional pilots.
3. Cohort Dashboards: Allowing teachers and bootcamp instructors to view evidence readiness scores across a whole class at a glance.

Built With

• next.js
• react
• typescript