Autonomous Bounty Hunter Agent

An AI agent that autonomously finds, evaluates, and claims GitHub bounty tasks.

How It Works

Search GitHub → Evaluate Issues → Comment/Claim → Write Code → Submit PR
     ↑                                                           |
     └───────────── Cron every 2 hours ◄─────────────────────────┘

Architecture

1. Scanner — Searches GitHub for bounty-labeled issues using gh CLI
2. Evaluator — Assesses issue difficulty vs agent capabilities
3. Commenter — Posts natural-language comments expressing interest
4. Implementer — Forks repo, writes code, runs tests
5. Submitter — Creates PR with /claim tag

Components

1. Issue Scanner (scanner.py)

Searches GitHub with multiple strategies:

• gh search issues "bounty" --state=open --language=python
• gh search issues "help wanted" --state=open --language=python
• Algora.io bounty board scraping
• Direct repo monitoring for new issues

2. Issue Evaluator (evaluator.py)

Scores each issue on:

• Language match (Python, JavaScript, TypeScript)
• Complexity (lines of code needed, domain knowledge required)
• Competition (how many people already attempted)
• Freshness (how recently created)
• Reward value

3. Autonomous Worker (worker.py)

Core loop:

async def work_cycle():
    issues = await scanner.find_bounties()
    for issue in issues:
        score = evaluator.score(issue)
        if score > THRESHOLD:
            if not issue.assignees:
                await commenter.express_interest(issue)
                if score > HIGH_THRESHOLD:
                    await implementer.start_work(issue)

4. PR Submitter (submitter.py)

• Clones repo via gh repo fork
• Creates feature branch
• Commits changes
• Pushes and creates PR with /claim #ISSUE
• Monitors for review comments and responds

Tech Stack

• Python 3.13+
• GitHub CLI (gh) for all GitHub operations
• httpx for API calls
• asyncio for concurrent operations
• AI model for code generation and natural comments

Usage

# Set up
export GITHUB_TOKEN="your-token"

# Run once
python worker.py --once

# Run continuously (every 2 hours)
python worker.py --interval 7200

Example Output

[2026-04-13 18:00] Scanner found 15 new bounty issues
[2026-04-13 18:01] Evaluator: rustchain-bounties#2957 scored 85/100
[2026-04-13 18:02] Commented on rustchain-bounties#2957
[2026-04-13 18:05] Started work on rustchain-bounties#2957
[2026-04-13 18:30] PR submitted: https://github.com/.../pull/3024
[2026-04-13 18:30] Claim tagged: /claim #2957

Results (First Day — 2026-04-13)

Action	Count
Issues evaluated	15+
Comments posted	7
PRs submitted	2
Code written	~1500 lines
Repos engaged	4

Why This Works

1. Volume — Evaluates more issues than a human could browse
2. Speed — Comments within minutes of issue creation
3. Quality — Focuses on issues matching actual capabilities
4. Persistence — Runs 24/7, never sleeps
5. Learning — Tracks which repos respond, prioritizes active maintainers