Not a bug report - triage data, in case it is useful for the open-PR queue.

I clustered all **2091 open PRs** by title, then verified every candidate against the files each PR actually changes. Below are only the groups where the PRs touch the **same primary file for the same stated purpose**. Tests, docs, READMEs and assets were excluded from the file comparison, since two PRs both adding a test file is not evidence of anything.

## 78 open PRs, 16 distinct fixes

| # | What | PRs |
|---|---|---|
| 10 | `omarchy-windows-vm` setgid / set-ID on mount sources | #9414, #9504, #9564, #9605, #10046, #10411, #10489, #10595, #10906 (+1) |
| 8 | Grok usage collector | #6485, #6902, #7200, #8280, #8649, #9498, #9840, #10074 |
| 8 | Antigravity usage collector | #7701, #8446, #8479, #8693, #8911, #9216, #10515, #10554 |
| 8 | `omarchy toggle bar` on/off inverted | #7023, #9074, #9102, #9455, #10099, #10623, #10635, #10932 |
| 7 | OpenCode usage collector | #6526, #7082, #7687, #7765, #8065, #8823, #9944 |
| 4 | Sunshine canonical systemd unit | #8136, #8879, #9104, #9979 |
| 4 | Idle-inhibit Steam games | #7334, #9651, #9667, #10447 |
| 4 | Universal clipboard on non-Latin layouts | #6419, #7285, #7375, #10247 |
| 4 | Combined dual-battery reporting | #7138, #7186, #10384, #10392 |
| 4 | Dismiss low-battery warning on charge | #7688, #8838, #8895, #10900 |
| 3 | Vivaldi browser support | #5774, #6593, #9430 |
| 3 | Fuzzy menu search | #7016, #7358, #8987 |
| 3 | Hermes usage collector | #6645, #6647, #9871 |
| 3 | Copilot usage collector | #7798, #8972, #9945 |
| 3 | Remember menu position when navigating back | #7074, #7800, #7954 |
| 2 | Sublime Text default editor (`subl`) | #9656, #9673 |

The `omarchy-windows-vm` group is the sharpest: all ten change **only** `bin/omarchy-windows-vm`, for the same setgid clearing. One of them (#9605) is omarchybot's own.

## The larger pattern behind it

**At least 67 open PRs touch `bin/omarchy-agent-usage-*`**, across 24 providers:

```
grok 11   codex 11   update 10   opencode 10   claude 9   antigravity 7
cursor 4   kimi 3   hermes 3   fireworks 3   copilot 3   agy 3
zai 1   synthetic 1   snapshot 1   pi 1   openrouter 1   openclaw 1
omp 1   ollama 1   muse 1   minimax 1   kiro 1   deepseek 1
```

Half of those providers appear exactly once, so they are not duplicates of anything — they are 24 people independently writing the same shape of script.

A further **10 PRs** add a provider to the coding-agent picker in `bin/omarchy-agent` (#6827, #6844, #6969, #7261, #7693, #8051, #8450, #9257, #10079, #10913) — those are not duplicates of each other, they add different agents.

Where the same provider appears many times over, that is duplication. Where many different providers each need a near-identical script plus SVG pair plus manifest entry plus scanner test, that reads more like a missing extension point than sixty separate contributions. That is a design call, not something triage data can settle.

## Method and limits

- Clustered titles by shared rare tokens, then **required a shared primary file** before calling anything a duplicate.
- Groups here were additionally checked by hand.
- **False positives the file check caught**, listed so the method's failure mode is visible: #9495 / #9348 look like duplicate ThinkPad speaker tuning but are `gen11` vs `gen6`, different directories; #9838 clusters with the menu-ranking PRs but adds `ConfirmDialog.qml`, a different feature; the `/tmp` hardening PRs (#10503-#10506) name different targets and are a deliberate series.
- The automated pass produced ~146 candidate groups covering ~349 PRs. I am deliberately **not** publishing that list, because the heuristic mislabels in both directions and the wider set has not been hand-checked. The 16 above have.
- Snapshot taken 2026-09-09.

Happy to share the full candidate list, the scripts, or re-run it whenever it would help. No action needed on this issue itself — close it whenever it has served its purpose.