sn-update

1# sn-update
2 
3Refresh installed `sn-*` skills from upstream
4[SenseNova-Skills](https://github.com/OpenSenseNova/SenseNova-Skills).
5 
6---
7 
8## Decide the scope
9 
10- No list given → every `sn-*` skill upstream.
11- Specific skills named → only those. Don't expand.
12- A named skill missing upstream → surface as error and stop.
13- User said "force / 强制" → re-install even when up-to-date.
14 
15---
16 
17## Decide the target agent
18 
19Check which directories exist:
20 
21| `~/.openclaw/skills/` | `~/.hermes/skills/` | Target |
22|---|---|---|
23| exists | absent | openclaw |
24| absent | exists | hermes |
25| exists | exists | **ask the user** — never silently dual-write |
26| absent | absent | no install found, stop |
27 
28---
29 
30## Sync the upstream repo
31 
32Persistent cache at `~/.cache/sn-update/repo/`. Default URL:
33`https://github.com/OpenSenseNova/SenseNova-Skills.git`. User may override
34with a fork URL.
35 
36- **First run**: if you want to actually limit blob download, use partial
37  clone with `--filter=blob:none --no-checkout`, then sparse-checkout only
38  the selected `skills/<name>` paths before copying them. `--filter=blob:none`
39  alone does **not** keep the cache small if the full worktree is checked out;
40  that checkout will still download most or all needed blobs. It still
41  preserves history metadata for SHA queries.
42- **Subsequent runs**: fetch + hard-reset to the upstream default branch, and
43  re-apply sparse-checkout for only the requested `skills/<name>` paths before
44…