BountyPilot AI — Autonomous Web3 Bounty & Work Verification Agent

Developed for the Devpost Learn Hackathon: Build With AI Basics (Challenge #30447)
Systematically planned and implemented using the official Devpost Learn Skill Pack (1-start, 2-scope, 3-prd, 4-spec, 5-build, 6-ship).

Inspiration

While thousands of bounties and freelance milestones are posted across decentralized developer marketplaces (WorkUSDC, Superteam, Farcaster), software builders routinely encounter three painful barriers:

1. Unfunded / Speculative Bounties: Postings that promise tokens but lack verifiable smart-contract escrow.
2. Subjective Acceptance Criteria: Vague milestone specifications that lead to uncompensated disputes.
3. Unprovable Proof-of-Work: No standard, tamper-proof cryptographic attestation linking the submitted code to the original work specification.

BountyPilot AI solves this by acting as an autonomous copilot that conducts real-time marketplace diligence, enforces strict payment compliance gates (PC0 to PC5), and produces verifiable SHA-256 execution attestations.

What It Does

• Multi-Venue Live Discovery: Queries real-time work marketplaces (including WorkUSDC REST API https://workusdc.com/api/tasks) to discover open developer tasks.
• Strict Policy Compliance Engine (PC0–PC5):
  • PC0: Verifies platform canonical identity.
  • PC1: Checks nominal advertised budget without fabricating missing amounts.
  • PC2: Enforces non-volatile settlement in USDC on Base Network (or verified Ethereum escrow).
  • PC3: Validates smart contract escrow attachment (0x...).
  • PC4: Audits specification depth and deliverable clarity.
  • PC5: Executes final diligence attestation.
• Cryptographic Proof Generator: Ingests raw task payloads, computes canonical SHA-256 hashes, binds the work to the authorized settlement destination (0xC6F86e170411182114FcCdb28793dC76B5e8D144), and exports an immutable JSON Proof Packet.
• Real-Time Telemetry Dashboard: A responsive, high-contrast local web interface (http://127.0.0.1:4040) for scanning, filtering, and 1-click proof packet generation.

How We Built It

In strict adherence to the Devpost Learn methodology, we planned the entire architecture before writing production code:

• devpost/learner-profile.md: Defined our flipped-interaction parameters and engineering focus.
• devpost/scope.md: Formulated the unique kernel (zero-trust cryptographic evidence generation) and strict POC boundaries.
• devpost/prd.md: Detailed the core journey, screen interactions, policy states, and non-goals.
• devpost/spec.md: Structured the modular TypeScript/Node.js architecture and data contracts.
• devpost/app-map.html: Built an interactive visual blueprint of component interactions.
• devpost/checklist.md: Maintained an exhaustive verification and testing checklist.

The engine was implemented in TypeScript with zero heavy dependencies, utilizing native Node.js HTTP and crypto primitives, accompanied by a 100% green unit test suite (6/6 passing).

Challenges We Overcame

1. Handling Live Network Degradation: Public decentralized APIs can experience rate limits or transient outages. We implemented an isolated fallback layer serving certified immutable fixtures when offline, ensuring the agent remains testable anywhere without faking false success.
2. Deterministic Canonical Hashing: Standardizing JSON payloads across platforms required strict key sorting and canonical utf-8 encoding before computing SHA-256 digests.

What We Learned

Planning before building with an AI coding agent fundamentally changes software quality. When the agent is grounded in an approved Scope, PRD, and Technical Spec, it avoids speculative patches, preserves core invariants, and implements clean, maintainable architectures on the first pass.

What's Next

• Integrating autonomous LLM milestone solver agents to draft initial code solutions.
• Autonomous on-chain escrow release triggers via verified smart contract calls on Base.
• Multi-channel notification alerts via Farcaster and Telegram.

Built With

• ai-agents
• base
• cryptography
• devpost-learn
• node.js
• rest-api
• typescript
• usdc
• web3