Codex Project Recovery Manager V2

Inspiration

What it does

How we built it

Challenges we ran into

Accomplishments that we're proud of

What we learned

What's next for Codex Project Recovery Manager V2

Inspiration

Returning to an older local software project often starts with uncertainty: is this the right folder, is documentation present, and what is the smallest safe next step?

Codex Project Recovery Manager V2 is a new standalone competition project inspired by a real project-management need from my earlier work. The earlier public repository is inspiration only; no private folders, credentials, or real project records were copied into V2.

What it does

Recovery Manager V2 is a Windows desktop app. It audits only names, directory structure, common project markers, Git-folder presence, entry-point names, and modification dates. It never opens file contents, secrets, credentials, or .env contents, and it never changes the selected project.

It shows evidence, a NOW/LATER/ARCHIVE priority, a four-stage Preserve, Understand, Verify, Continue plan, and a bounded safe prompt for Codex. Markdown export is user-controlled and omits absolute local paths by default.

How I built it

I built the standalone Python and Tkinter application with Codex and GPT-5.6. Codex assisted with implementation, tests, documentation, and local verification within my safety constraints. The project includes two invented offline demos and unit tests.

Challenges

The challenge was producing useful recovery guidance without reading project contents. Missing markers are treated as evidence, not certainty. Another challenge was preventing private information from appearing in reports or prompts; tests verify secret-content exclusion and private-path omission.

Accomplishments

Runnable Windows desktop app with no third-party dependency
Read-only metadata audit and four-stage Recovery Plan
Bounded safety-constrained Codex prompt
User-controlled Markdown export
Two offline synthetic demos and automated tests

What I learned

Recovery tooling can provide value before permission to edit. Carefully selected metadata can propose a useful first step, while explicit constraints make a later AI handoff easier to review.