| name | files |
| description | Upload, list, delete, and get download URLs for files in Kagura Memory's R2 storage (sha256 integrity-bound; an upload can optionally be bound to an owning context for access control). Use when the user wants to attach or manage raw files in their workspace. |
kagura files — R2 file storage
Drive the kagura files command group (presigned-PUT uploads with sha256
binding). Thin wrapper around the installed kagura CLI.
Preflight
kagura --version; if missing → uv tool install kagura-memory (or
pip install kagura-memory), then stop.
- Requires authentication. Run
kagura auth status; if not authed, run the
auth skill first.
Run (choose by intent)
kagura files upload ./data.bin
kagura files upload ./data.bin --binding-context-id <ctx>
kagura files list
kagura files download-url <file_id> -c <ctx>
kagura files delete <file_id> -c <ctx>
Consume the result
- Relay the returned file id and download URL. Download URLs are short-lived
— do not cache them; re-run
download-url when one expires.
--binding-context-id (the file's owning context for ACL) is distinct from
--context-id (the workspace). Bound files route read/write/list/delete through
that context's ACL; an unbound upload stays workspace-scoped. A denied download
is reported as a 404 (existence-hiding), not a 403.