An evidence-first maintenance skill for Codex and Claude Code that detects metadata drift across manifests, lockfiles, documentation, deployment configuration, and catalogs, then aligns local files only when requested.
fix-project-metadata-mismatches is a AI-agent skill created to make a real workflow clearer, faster, and easier to trust. It turns a focused product idea into something people can use, evaluate, and understand without needing to inspect the source code.
Why fix-project-metadata-mismatches exists
Project identity and version details are duplicated across files and drift, leaving users with broken links, wrong install commands, or conflicting compatibility claims.
The goal is not simply to add another tool. fix-project-metadata-mismatches is designed to reduce the friction around that problem and give people a more dependable way to complete the work.
What it delivers
- Agents can identify conflicting names/versions/URLs.
- Repair authorized local copies.
- Preserve unknowns for manual resolution.
- Report remaining remote actions separately.
The experience
Audits and optionally synchronizes project identity and release metadata across authoritative local files.
The project brings the important parts of the workflow into one coherent experience. Its interface and behavior are shaped around the people using it, while the implementation stays focused on predictable results and honest feedback.
What to know
Every product has boundaries. These are the most important considerations for fix-project-metadata-mismatches:
- Each field has a different authoritative source.
- A human-facing description must match implemented behavior, not merely manifest prose.
- Local synchronization does not authorize publishing or changing remote metadata.
These limitations are presented openly so visitors can understand where the project works well, what it depends on, and what may change in future versions.
Current status
Published npm package v0.1.0. The current release demonstrates the core value of the project and the main end-to-end experience.
My role was agent workflow design and automation engineering, covering agent workflow, validation automation, and packaging. The related case study explains the decisions and delivery process in greater depth for visitors who want the engineering story behind the product.