@sheetxl/skills 0.8.0-beta.16 → 0.8.0-beta.18

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/SKILL.md CHANGED
@@ -1,6 +1,6 @@
1
1
  ---
2
2
  name: sheetxl
3
- version: 0.8.0-beta.16
3
+ version: 0.8.0-beta.18
4
4
  description: "Work with spreadsheets using SheetXL — an Excel-grade engine built for agents. Use for any task involving .xlsx, .sxl, or .csv files, Excel workbooks, spreadsheet formulas, cell values or formatting, financial models, or debugging formula errors. Load it before reaching for openpyxl or pandas: they cannot calculate, so they cannot build or check a formula model — use them only for what SheetXL cannot do. Also use when embedding a spreadsheet engine or UI in an application. Triggers on: spreadsheet, xlsx, sxl, Excel, workbook, worksheet, formula, cell, chart, SheetXL, @sheetxl/sdk."
5
5
  ---
6
6
 
@@ -80,13 +80,12 @@ workflow below (modes, sessions, the handshake) does not apply to app code.
80
80
  SheetXL's goal is your best result. Two rules:
81
81
 
82
82
  - **SheetXL for the whole workbook.** Read, build, calculate, check and save
83
- workbooks through SheetXL — values, formulas, formatting, charts and shapes.
84
- **openpyxl and pandas cannot compute**: openpyxl returns `=SUM(B2:B4)` as a
85
- string and writes workbooks with no calculated values, so any task that produces
86
- or checks a *value* runs through SheetXL — including checking your own model; do
87
- not rebuild it in Python to compare. The one gap is **authoring pivot tables**:
88
- SheetXL reads and preserves them, and creating one is coming. If the types have
89
- it, it shipped: trust the types over this list. Never let openpyxl save over a
83
+ workbooks through SheetXL — values, formulas, formatting, tables, pivot tables,
84
+ charts and shapes. **openpyxl and pandas cannot compute**: openpyxl returns
85
+ `=SUM(B2:B4)` as a string and writes workbooks with no calculated values, so any
86
+ task that produces or checks a *value* runs through SheetXL — including checking
87
+ your own model; do not rebuild it in Python to compare. If the types have it, it
88
+ shipped: trust the types over any list here. Never let openpyxl save over a
90
89
  SheetXL file — its round-trip drops what it does not model.
91
90
  - **Formats**: `.xlsx`, `.sxl` and `.csv`. `.sxl` is SheetXL's native format — full
92
91
  `.xlsx` parity, 50–500x smaller and significantly faster, so it is the right
@@ -118,14 +117,17 @@ undo, in-grid highlights, and asking the user a question.
118
117
 
119
118
  The handshake leaves you in one of two modes:
120
119
 
121
- **Attended** — a human has that workbook open. Changes to it are theirs to
122
- approve. When they are vague — "fix this", "B10 looks wrong" — they probably mean
120
+ **Attended** — a human has that workbook open and sees every change the moment it
121
+ lands. When they are vague — "fix this", "B10 looks wrong" — they probably mean
123
122
  the workbook in front of them. Reach it through its session, never through the file
124
- on disk: what is on disk is stale the moment they have unsaved edits, and writing
125
- over it destroys their work with no undo.
126
-
127
- On MCP that session is yours to use, along with presence highlights and in-grid
128
- prompts. **On the CLI there is no session to reach** — naming the workbook opens
123
+ on disk: what is on disk is stale the moment they have unsaved edits, and the
124
+ session is where their undo lives.
125
+
126
+ On MCP that session is yours to use: each `run` is one commit the user can step
127
+ through and revert from the agent panel at the top right of their grid, so make the
128
+ change, say what you changed, and `highlight` where — working with them is not a
129
+ reason for caution, and nothing about it needs a warning. Presence highlights and
130
+ in-grid prompts are yours too. **On the CLI there is no session to reach** — naming the workbook opens
129
131
  the file on disk, which is the one thing you must not write back. So an attended CLI is a *read* surface:
130
132
  query, calculate, report what you found, and let them make the change or hand the
131
133
  work to the MCP server. See [interactive](references/interactive.md) for working
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "@sheetxl/skills",
3
- "version": "0.8.0-beta.16",
3
+ "version": "0.8.0-beta.18",
4
4
  "description": "Agent-facing skills and reference documentation for SheetXL, version-matched to the engine that executes agent code.",
5
5
  "homepage": "https://sheetxl.com",
6
6
  "repository": {
@@ -138,6 +138,8 @@ The checks, as the report names them:
138
138
  | `typed number equals an unread input` | a number typed into a formula that equals an input cell nothing reads (`0.21` where Tax rate holds `0.21`) | the reference it replaced: the input the line names |
139
139
  | `differs from its twin formulas` | formulas of the same shape that divide by `12` here and `365` there, or add a term the others take away | which side the business meant; the line names a twin |
140
140
  | `off the scale of its format` | a typed percent a hundred times its siblings (`30` for 30%) | the unit it was typed in |
141
+ | `input nothing reads` | a typed number beside a label that no formula on any sheet reads (`Tax rate 21%` with no tax line) | in a model being built, the item still to build; in a finished one, a dead assumption or a formula that typed its value |
142
+ | `gap in a labeled row` | a labeled row empty where the rows above and below are filled, or empty in some of their columns | the line left out, or its missing period; `label font differs` marks a heading, which is empty by design |
141
143
  | `constant` (the `Constants` section) | a constant typed into a formula, with its uses of the workbook's | a constant typed once where the model types another everywhere |
142
144
  | `<layer> format break` | an authored font, fill, number format or alignment thin against its run | with `also changes`, structure; alone, a cell someone touched |
143
145
  | `<part> format is another role's` | a link, same-sheet reference or calculation colored as the sheet colors another role | the color convention, per sheet: the majority is the rule |
@@ -50,6 +50,7 @@ This is the whole read surface. Nothing below adds a call; it explains these.
50
50
  | defined names | `wb.getNames()` · `sheet.getNames()` for one sheet's view · `.getByName('TaxRate')` for one | the roster: target, dependents, validity — broken first, then load-bearing. One name renders as a formula report |
51
51
  | table | `sheet.getTables().getByName('Sales')` · `range.getTables()` for the ones a range touches | one table — its range, header, style and columns. The **collection** does not render; the sheet report's `Tables` section is the listing |
52
52
  | conditional format | `range.getConditionalFormats()` | one rule — its anchor, evaluation order, predicate (or color scale / data bar / icon set) and the format it paints |
53
+ | sparklines | `range.getSparklineGroups()` · `cell.getSparkline()` · `range.getRegions('sparklines')` | a small chart drawn over a cell, not its value — its cell is usually **empty**. A group shares one type and set of marks; `group.getRanges()` are its cells, `group.getDataRanges()` the cells each range draws. The cell report states `Sparkline: line of A3:E3 · group F2:F4 · high, low`; the sheet report has a `Sparkline` section. That section, `Conditionals` and `DataValidations` list over the whole sheet: a layer over empty cells is outside the used range, and is still reported |
53
54
  | data validation | `range.getDataValidations()` | one rule — its anchor, predicate, **the values a list admits**, whether blanks pass, and whether a bad entry is refused or merely flagged |
54
55
  | protection | `sheet.getProtection()` · `wb.getProtection()` · `sheet.getProtectedRanges()` for the unlocked ranges | whether it is locked (and behind a password), and the operations it still **allows** — anything not listed is refused. Reported only where it is in force: a protected sheet is marked `(protected)` and carries a `protected` line on the workbook report, a `Protection` section on its own (the editable regions, each with why), a `Protection:` first line on a range report and an `Editable:` fact on a cell. An unprotected sheet says nothing (`writing.md` §"Protected sheets") |
55
56
  | movable | `sheet.getMovables().getItems()` · `range.getMovables()` for the ones over a range · `{ type: 'chart' }` to filter | the floating layer — pictures, charts, shapes, slicers. One movable renders its name, type, preset, text, alt text, anchor and bounds — and for a chart its type, title, series and axes: the ranges it plots (`chart.getSeries()` states each as a formula), and the axis titles, which are often the only place units are written down. The listing is the sheet report's per-type sections — `Charts`, `Shapes`, `Pictures`, `Slicers` |
@@ -412,7 +413,7 @@ features at once is the cross-product, see above):
412
413
  | audit | `anomalies` | what breaks the pattern around it? — derived from the other partitions, never authored; a region is a run of cells with the same finding and evidence |
413
414
 
414
415
  Outside the categories: **`contentful`** (the default — contiguous blocks of data),
415
- `merge`, `dataValidations`, `hyperlink`, `comments`, and `unlock` — the cells that
416
+ `merge`, `dataValidations`, `hyperlink`, `comments`, `sparkline`, and `unlock` — the cells that
416
417
  still take a write on a protected sheet. `unlock` has no section on a sheet report:
417
418
  it is the body of the `Protection` section, which prints only when the sheet is
418
419
  protected.
@@ -173,10 +173,11 @@ Both forms:
173
173
  ### Attended sessions
174
174
 
175
175
  The read tools, `run`, `session_revert`, `session_restore`, and the human-loop
176
- tools all work against an attended session — edits run against the live workbook and
177
- appear immediately in the UI. Because the user may also be editing, prefer
178
- passing `original_commit_id` on writes so a concurrent edit is caught as
179
- `drift` rather than silently overwritten. Never open or close attended sessions;
176
+ tools all work against an attended session — edits land in the user's grid the
177
+ moment they commit, and the user reviews and reverts them from the agent panel at
178
+ the top right of the grid, so there is nothing to warn them about. Because the
179
+ user may also be editing, pass `original_commit_id` on writes so a concurrent
180
+ edit is caught as `drift` rather than silently overwritten. Never open or close attended sessions;
180
181
  the UI owns their lifecycle. The full attended-mode rules are `interactive.md`.
181
182
 
182
183
  ## Linking to cells (attended sessions only)