feat: add autonomous bounty-hunting agent with multi-LLM orchestration (#861) [fj4WqyCCw3C5ShR1RfB7MoBPTpkRrBFYP1uT35g3MvT]

[waterWang]

Summary

Implements a full autonomous multi-agent system that finds bounties, analyzes requirements, implements solutions, runs tests, and submits PRs without human intervention, as specified in the #861 T3 bounty.

Multi-LLM Agent Pipeline

Role	Agent	Description
Scanner	BountyHunter.scan()	Discovers unclaimed bounty issues via GitHub API
Planner	PlannerAgent	Analyzes requirements → structured implementation plan
Implementer	ImplementerAgent	Generates TypeScript code following SDK conventions
Verifier	VerifierAgent	Runs tests, type-checks, linting + static analysis (TODO/console/secret detection)
Reviewer	ReviewerAgent	Multi-LLM scoring (quality, correctness, security)
Submitter	SubmitterAgent	Creates PR with wallet address for bounty payment

New Files

• sdk/src/agent/types.ts — Agent system type definitions (roles, plan, review, verification)
• sdk/src/agent/agent.ts — Base LLM agent class with provider abstraction + mock mode
• sdk/src/agent/planner.ts — Planning agent
• sdk/src/agent/implementer.ts — Code generation agent
• sdk/src/agent/verifier.ts — Test/verification agent with static analysis
• sdk/src/agent/reviewer.ts — Multi-LLM review agent
• sdk/src/agent/submitter.ts — PR submission agent
• sdk/src/agent/orchestrator.ts — BountyHunter orchestrator tying the pipeline together
• sdk/src/agent/index.ts — Module exports
• sdk/src/__tests__/agent.test.ts — 13 tests covering the full agent system

Configuration

import { BountyHunter } from '@solfoundry/sdk/agent';

const hunter = new BountyHunter({
  defaultProvider: { provider: 'openai', apiKey: 'sk-...', model: 'gpt-4o' },
  github: { owner: 'SolFoundry', repo: 'solfoundry', token: 'ghp_...' },
  walletAddress: 'fj4WqyCCw3C5ShR1RfB7MoBPTpkRrBFYP1uT35g3MvT',
});

const result = await hunter.hunt(861);
console.log(result.prUrl);

Testing

216 tests pass across all 11 SDK test files (including 13 new agent tests). Uses mock mode for deterministic LLM responses in tests.

Closes #861