Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Guide for writing Neovim plugins in Lua following official Neovim conventions (https://neovim.io/doc/user/lua-plugin/). Use this skill whenever the user is creating, modifying, or reviewing a Neovim plugin — including when they mention plugin structure, ftplugin, health checks, keymaps, setup() functions, vimdoc, LuaCATS annotations, or lazy loading in the context of Neovim plugin development. Also trigger when the user is working in a directory that looks like a Neovim plugin (contains plugin/,
.claude/skills/fredrikaverpil-nvim-plugin/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -12% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -48% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -41% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -65% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -39% | 0% |
:h lua-plugin (https://neovim.io/doc/user/lua-plugin/) is the authority on layout, setup() patterns, <Plug> mappings, guard variables, health checks and deprecation. Read it rather than working from memory — it is opinionated and it changes. This skill only records what it leaves out.
Per .stylua.toml: 2-space indentation, double quotes, 120-char width, requires sorted by stylua.
panvimdoc, then :helptagsdoc/.
lua-typecheck-action in CI, so luals catches breakage before users do.
luarocks-tag-release or release-please-action.Publish to luarocks when the plugin has Lua dependencies or is one itself.
:restart reloads plugin changes.nvim --startuptime /tmp/nvim-startup.log to check what eager loading inplugin/ costs.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 18,387 | 13,924 | -24% | 1 | 1 | 0% | 3,358 | 2,943 | -12% | 0 | 0 | — |
case-02 | fail→pass | 9,335 | 3,593 | -62% | 1 | 1 | 0% | 1,769 | 917 | -48% | 0 | 0 | — |
case-03 | fail→pass | 13,263 | 6,994 | -47% | 1 | 1 | 0% | 2,431 | 1,440 | -41% | 0 | 0 | — |
case-04 | pass→pass | 8,992 | 3,117 | -65% | 1 | 1 | 0% | 1,615 | 850 | -47% | 0 | 0 | — |
case-05 | fail→pass | 14,152 | 3,517 | -75% | 1 | 1 | 0% | 2,464 | 873 | -65% | 0 | 0 | — |
case-06 | pass→pass | 12,769 | 4,383 | -66% | 1 | 1 | 0% | 2,018 | 933 | -54% | 0 | 0 | — |
case-07 | pass→pass | 16,442 | 7,174 | -56% | 1 | 1 | 0% | 2,747 | 1,553 | -43% | 0 | 0 | — |
case-08 | fail→pass | 13,583 | 7,615 | -44% | 1 | 1 | 0% | 2,272 | 1,389 | -39% | 0 | 0 | — |
case-09 | pass→pass | 9,393 | 3,171 | -66% | 1 | 1 | 0% | 1,511 | 751 | -50% | 0 | 0 | — |
case-10 | fail→pass | 12,063 | 2,012 | -83% | 1 | 1 | 0% | 1,946 | 625 | -68% | 0 | 0 | — |
case-11 | pass→pass | 3,147 | 2,133 | -32% | 1 | 1 | 0% | 451 | 571 | +27% | 0 | 0 | — |
case-12 | pass→pass | 5,950 | 2,297 | -61% | 1 | 1 | 0% | 997 | 624 | -37% | 0 | 0 | — |
case-13 | fail→pass | 8,567 | 1,628 | -81% | 1 | 1 | 0% | 1,452 | 457 | -69% | 0 | 0 | — |
case-14 | fail→pass | 12,591 | 2,235 | -82% | 1 | 1 | 0% | 2,077 | 581 | -72% | 0 | 0 | — |
case-15 | pass→pass | 29,270 | 2,069 | -93% | 1 | 1 | 0% | 2,220 | 534 | -76% | 0 | 0 | — |
case-16 | fail→pass | 17,306 | 7,148 | -59% | 1 | 1 | 0% | 2,800 | 1,439 | -49% | 0 | 0 | — |
case-17 | pass→pass | 10,256 | 8,668 | -15% | 1 | 1 | 0% | 1,738 | 1,741 | +0% | 0 | 0 | — |
case-18 | pass→pass | 10,796 | 1,600 | -85% | 1 | 1 | 0% | 1,698 | 487 | -71% | 0 | 0 | — |
case-19 | fail→pass | 7,701 | 1,781 | -77% | 1 | 1 | 0% | 1,272 | 544 | -57% | 0 | 0 | — |
case-20 | pass→pass | 9,192 | 8,010 | -13% | 1 | 1 | 0% | 1,580 | 1,621 | +3% | 0 | 0 | — |
case-21 | pass→pass | 12,163 | 8,883 | -27% | 1 | 1 | 0% | 2,186 | 1,936 | -11% | 0 | 0 | — |
case-22 | pass→pass | 15,903 | 12,238 | -23% | 1 | 1 | 0% | 2,587 | 2,415 | -7% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +45 percentage points is the difference between those two pass rates over the 22 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.