139 — Umbra: concrete encrypted multipart paste storage
Checked 2026-09-05 UTC. Implementation analogue; no XZ writer identified.

https://github.com/henomis/umbra
Public GitHub metadata records creation December28,2025 and last push January5,2026. The captured tree is f485736b317de5635a722ee7abcacf4d424a2372. Six source files covering encryption, content, manifest, upload, provider selection and configuration were fetched through GitHub blob API and verified against Git blob SHA1. Private139 records files and SHA256 hashes. No code was run or data uploaded.

The inspected implementation splits files, encrypts chunks, uploads redundant copies to configured providers and records their metadata. Its cipher is XChaCha20-Poly1305 with Argon2id password derivation. Its separate manifest includes magic8690998b, version1, cipher/KDF identifiers, salt and nonce before encrypted content. Provider selection supports termbin, clbin, pipfi and pastecnetorg; no Chinese Ubuntu adapter is listed. Chunk size is configurable rather than an identified30000-character naming convention.

This does not match the observed large XZ objects' nested Fernet layout. No claim is made that every possible wrapper or fork is excluded. The useful distinction is architectural: a small encrypted manifest can index larger uploaded objects while the payload and manifest follow different serialization paths. Establishing that role for xz_knowledge still requires a linked reader or producer; this example cannot identify its plaintext.

Searches combining Fernet/chunk/pastebin and the exact Chinese host with splitting or Base64 terms did not surface a matching writer. They led to Umbra through the general encrypted-file-storage query. The source inspection materially narrows that candidate to an analogy.