# AI Output Markers

Scriba adds an AI-generated marker to converted files to support legal and contractual transparency requirements.

## What gets added

Each converted file includes structured metadata fields:

- `@scriba-ai-generated: true`
- `@scriba-marker-version: 1`
- `@scriba-source-language: ...`
- `@scriba-target-language: ...`
- `@scriba-conversion-id: ...`
- `@scriba-timestamp: ...`
- `@scriba-platform-version: ...`

## Format by language

- C-family: `/* ... */`
- Python / Ruby / Shell: `# ...`
- SQL / Ada style: `-- ...`
- Erlang / Prolog style: `% ...`
- Lisp / Clojure style: `; ...`

For strict non-comment formats (for example JSON/XML/HTML), Scriba creates a sibling marker file:

`<filename>.scriba-marker.json`

## Placement rules

- Marker is inserted at the top of file content.
- For shebang files (`#!/...`), marker is inserted on line 2.
- Header-aware behavior also handles XML declarations, encoding headers, and preserved license headers.

## Customer responsibilities

- Preserve marker comments/sidecar files in downstream workflows.
- Do not remove marker metadata from redistributed generated outputs.
- If your delivery pipeline modifies generated files, ensure marker presence remains verifiable.