> ## Documentation Index
> Fetch the complete documentation index at: https://docs.svrnos.com/llms.txt
> Use this file to discover all available pages before exploring further.

# How to contribute

> Propose a code, refine a definition, or suggest a clarification.

The SVRNOS Governance Error Register is a curated open standard. Contributions are welcome.

## What we accept

* **New code proposals** — a structural failure mode the existing 110 codes don't capture
* **Code refinements** — a sharper definition, a missing sub-type, a corrected "Distinct From" boundary
* **Documented events** — a real-world case that should be linked to an existing code
* **Translations** — a non-English rendering of a code's definition
* **Tag refinements** — a more precise harm/vector/sector tag for a code

## What we don't accept

* **Event reports.** SVRNOS does not host an event corpus. We map against the [AI Incident Database](https://incidentdatabase.ai). Report events there — we'll inherit the mapping.
* **Speculative codes** without at least one illustrative scenario from real-world observation.

## How to submit

Email **[hello@svrnos.com](mailto:hello@svrnos.com)** or use the [SVRNOS /ask channel](https://svrnos.com/ask).

For new code proposals, include:

1. **Proposed code number** (suggest one in the unassigned HTTP slots; we may renumber)
2. **Name** — short, distinctive
3. **One-sentence definition**
4. **Why this is distinct** from the closest existing codes — name them and explain the boundary
5. **At least one illustrative or documented case** (AIID ID if it exists)
6. **The structural property the code captures** — what failed, not what was generated

## Credit

Accepted contributions earn an entry in the [Attributions](/attributions) page. Contributors with formally adopted framings appear in the Zenodo DOI metadata for that GER version.

## The curation process

Sushee Nzeutem (SVRNOS) reviews every submission. Decisions follow these gates:

1. **Structural distinctness** — does this fail in a way no existing code captures?
2. **Regulator test** — can a regulator read this code name and immediately understand the failure?
3. **Evidence sufficiency** — is there enough documented or illustrative material to defend the code under scrutiny?

Rejected proposals get a written explanation. Accepted proposals land in the next minor version release.

## Pattern: this is MITRE CVE, not Wikipedia

The GER is curated, not crowd-edited. Public read, controlled write. Same model as MITRE CVE, IANA HTTP status codes, and MITRE ATT\&CK. This is deliberate — it's what makes the codes citable in court, in regulation, in insurance underwriting.
