I built a read-only tool for extracting boundary evidence from GitHub Issues — looking for real OSS cases

Many GitHub Issues are difficult before anyone touches the code.

The first challenge is often figuring out where the problem belongs:

• Is it caused by the current project or a dependency?
• Is there an upstream issue?
• Is it specific to a runtime or environment?
• Did it appear only after a particular version or change?
• Is the reported behavior actually a regression?

The relevant evidence is often scattered across the issue description and many comments, so I built Issue Boundary Evidence (IBE) to collect it into one report.

What IBE does

IBE takes a public GitHub Issue URL and extracts evidence related to:

• dependencies
• upstream and downstream boundaries
• runtime and environment
• versions
• regressions

It produces a Markdown report with links back to the original evidence.

What it does not do

IBE is deterministic and read-only.

It does not:

• use an LLM
• execute code found in the issue
• require access to a local repository
• comment on, close, or modify the original issue
• claim to determine the root cause

Its job is only to organize boundary evidence for human review.

Why I am posting this

I have tested IBE against historical public issues, but it has not yet been used by a confirmed external maintainer or developer.

I am now looking for real OSS cases.

If you maintain or contribute to an open-source project, you can either:

1. run IBE on a public GitHub Issue you are currently handling or genuinely care about; or
2. send me that Issue URL and ask me to analyze it.

Negative results are especially welcome.

For example:

• it identifies the wrong boundary
• it misses obvious maintainer evidence
• it mistakes a fork or example project for an upstream dependency
• version or regression signals are too noisy
• the report contains too much information
• the output is simply not useful in a real triage workflow

A failed case would be more valuable at this stage than generic encouragement.

Try it

python -m pip install issue-boundary-evidence

GitHub repository:

https://github.com/LaimaWu/issue-boundary-evidence

You can also submit a real-world test case through the repository’s issue template:

https://github.com/LaimaWu/issue-boundary-evidence/issues/new?template=real_world_issue_test.md

If you try it, please tell me what helped, what failed, and whether the output changed anything about how you understood the issue.

Disclosure: This post was drafted with AI assistance and reviewed and edited by the author before publication.