3 min read

I built a bot that reviews every PR on my open-source project

How a real need in StepFi became a Python project combining deterministic checks, security reviews, issue reminders, and Telegram notifications.

By Emmanuel Itighise

StepFi audit bot workflow showing Python code, GitHub pull request checks, an automated security review, and Telegram notifications.

It's been a while since I posted about my journey at Learn2Earn NG and a lot has happened. For context, Learn2Earn is an intensive coding fellowship programme training African developers through hands-on, project-based learning. Zero tuition, real-world projects, peer-driven model. Recently we had a learning pivot to Python and instead of just doing exercises I decided to build something real with it. The result is stepfi-audit-bot, an automated PR security auditor for StepFi, my open-source BNPL protocol on Stellar/Soroban.

The problem behind the bot

Here's the problem it solves. StepFi has multiple repos (API, Contracts, Frontend) and contributors submit pull requests regularly. Manually reviewing every PR for security gaps, CI tampering, vague descriptions, and code-to-issue alignment takes time I don't always have. Missed reviews mean merged vulnerabilities and for a fintech project that handles payment flows, that is not acceptable. So I built a bot that does it automatically.

How the review flow works

The core flow: every 2 minutes the bot polls StepFi-API and StepFi-Contracts repos for new or updated PRs. It fetches the linked issue, the PR description, every file diff, and CI status. Then it runs deterministic checks first, things like PR title relevance, description quality, merge conflict detection, and CI config tampering (flagging any modifications to workflow or test configs). After those checks pass, everything gets sent to an LLM for a security-focused code review. The bot then posts a GitHub review with a verdict: solves, partial, or does_not_solve, and auto-approves if the PR fully addresses the linked issue. Results also get reported to Telegram.

Following up on assigned issues

It also has a reminder system that tracks assigned issues with no PR linked. It escalates at 4h, 8h, 12h, 20h, 24h, 36h, and 40h intervals, posting GitHub comments and Telegram notifications so nothing slips through the cracks. Building this in Python made the whole thing click faster. Python's ecosystem for GitHub API interaction, HTTP requests, threading, and Telegram bot integration is mature and well documented.

Learn by solving a real problem

The Learn2Earn pivot to Python came at the right time because learning sticks faster when you are solving a real problem with it. I wasn't writing Python exercises, I was building a tool my project actually needed.

If you are learning a new language, my advice is to find a real problem in your existing work and solve it with that language. The syntax sticks when it matters.

Keep Learn2Earn NG in your tabs because the team is really cooking something great for African developers. Learn2Earn NG

About the author

Emmanuel Itighise

Emmanuel Itighise

Founder of Stepfi-App, @Pulsar-Stellar/sdk, and Trybe.

Emmanuel Itighise, known online as Emeditweb, is a full-stack and AI-native developer based in Lagos, Nigeria. He builds practical products and developer tools, with a focus on web applications, AI, and the Stellar ecosystem. He shares lessons from building in public, contributing to open source, and turning ideas into working software.