GitHub Issue Scout for Paid-Work Triage

This is a small Crawlee demo for finding public GitHub issues that may be worth human review before doing paid open-source work.

The goal is not to auto-comment or auto-claim bounties. The crawler collects public GitHub issue search results, extracts paid-work signals, flags risky patterns, and writes a ranked JSON shortlist that a human can inspect.

It was built as a working-code companion for an Apify/Crawlee technical article pitch.

Related writing assets:

• Portfolio page
• Paid services
• Case study
• Outreach kit
• Actor Store listing packet
• Apify current CFP alignment
• Apify submission packet
• Article draft
• Article pitch

Why This Demo Exists

Public bounty searches are noisy. A result can look valuable but still be a bad target because it is maintainer-only, core-team-only, already crowded, star-gated, or full of low-quality AI attempts.

This demo turns that lesson into a practical workflow:

• Collect issue candidates with Crawlee.
• Keep only structured public metadata.
• Score positive signals like reward text, helpful labels, and low comment count.
• Flag risk signals before spending engineering time.
• Produce a reviewable JSON artifact instead of posting spam.

Run It

npm install
npm run demo

The default output is:

output/leads.json

You can provide your own query:

npm start -- --query 'bounty "good first issue" state:open comments:<10 updated:>2026-04-01' --limit 10 --output output/custom-leads.json

Tip: add is:issue to GitHub search queries when you want to exclude pull requests from the result set.

Run as an Apify Actor Locally

The repository also includes an Apify Actor adapter:

• .actor/actor.json
• .actor/input_schema.json
• Dockerfile
• src/actor.ts

Run the Actor locally:

npm install
npm run actor:local

Local Actor output is written to Apify-compatible storage:

storage/key_value_stores/default/OUTPUT.json
storage/datasets/default/*.json

The Actor version accepts:

• queries: GitHub issue search queries
• limit: max results per query, capped at 30
• outputKey: key-value store key for the summary object

Output Shape

Each lead includes:

• repo
• title
• url
• comments
• labels
• score
• rewardSignal
• positiveSignals
• riskSignals
• verdict

Verdicts:

• pursue: worth opening manually and reproducing locally
• review: unclear, inspect before touching code
• skip: likely poor expected value

Safety Rules

• Do not auto-comment on GitHub issues.
• Do not claim a bounty until you have reproduced the issue and checked maintainer eligibility.
• Skip issues that are maintainer-only, core-team-only, star-gated, or already crowded.
• Treat this as a lead-scoring assistant, not an autonomous bounty hunter.

Article Angle

This demo is designed as the working example for an Apify/Crawlee technical article:

Building a GitHub Issue Scout with Crawlee and Human-in-the-Loop Triage

The article would walk through building the crawler, designing the scoring rules, and avoiding the common trap of wasting time on noisy or low-trust bounty issues.

Because it is Actor-ready, the article can also include a final section on turning the script into a reusable Apify Actor.