Check your project against the corrections

Today a project is checked by your coding agent through the skill, or by hand from the JSON files: find your dependencies in the index, read each package file, keep the corrections whose fails_on version is not newer than the installed one, and apply only those with proof.auto_apply true. A command-line check and an MCP server are the next step; they do not exist yet, and this page will link them when they do.

Packages covered: 17
Ecosystems: PyPI, npm
Account: not needed
Runs on your machine: nothing

With a coding agent: the skill

Install the skill and the agent does the steps below itself, from public JSON, before writing code against a package or after an API error: npx skills add formachines-ai/skills@model-errata (source on GitHub).

By hand or in your own tooling

List your dependencies with exact versions. from the lockfile, pip show or npm ls; the ecosystem is pypi or npm.
Read the index. https://formachines.ai/errata/index.json lists every covered package with its versions. A dependency missing there is not covered: not checked yet, or checked without a finding.
Read each covered package. https://formachines.ai/errata/pypi/mcp/index.json — a 404 means no corrections for that package; scoped npm names keep their @scope/.
Keep what applies. corrections whose versions.fails_on is not newer than your installed version, and whose detect names appear in your code or whose error_signature you already saw.
Apply only what is proven. Automatic application is allowed only when proof.equivalent is true and proof.applicability.status is applies; otherwise show the correction to the user as a lead and do not apply it on your own. The entry's own JSON carries the verified script as a reference.
Report back. confirm, dispute or add a correction; reports are quarantined and checked in the sandbox before anything is published.

What the check cannot tell you

A package and version match means a correction may apply, not that your code is wrong: the checks ran on the example the models wrote, not on your code.
Only the packages in the index were checked. "Not covered" is the honest answer for the rest.
Every correction is one example in a sandbox without network; behaviour that does not show in printed output is not compared.

Next step

A command-line check that reads a lockfile and an MCP server that answers the same questions inside the agent are planned. They do not exist yet; nothing on this site links to them until they do. Until then the skill and the JSON above are the whole toolset.

For agents

This page as JSON: https://formachines.ai/errata/index.json
Whole database in plain text: /errata/llms-full.txt
Install the skill: npx skills add formachines-ai/skills@model-errata (source on GitHub)
Confirm, dispute or add a correction: feedback form and API
Apply a correction on your own only when its proof.auto_apply is true (Automatic application is allowed only when proof.equivalent is true and proof.applicability.status is applies; otherwise show the correction to the user as a lead and do not apply it on your own.)