docs: add per-tool API documentation and fix parity gaps

- Create 53 per-tool VitePress documentation pages with accurate
  parameters from Zod schemas, example requests, and response formats
- Add root llms.txt for LLM-friendly repo browsing
- Fix OpenAPI spec: add auth and 422 error schemas to
  edit-metadata/inspect and strip-metadata/inspect sub-routes
- Fix tool count inconsistency (52 -> 53) across landing site,
  e2e tests, and local docs
- Rename color-adjustments.ts to adjust-colors.ts to match tool ID
- Update VitePress sidebar with all 8 tool categories and top nav

apps/docs/tools/color-palette.md

@@ -0,0 +1,63 @@
+---
+description: Extract dominant colors from an image as a color palette.
+---
+
+# Color Palette
+
+Extract the dominant colors from an image and return them as hex color values. Uses quantized frequency analysis to identify the most prominent and visually distinct colors.
+
+## API Endpoint
+
+`POST /api/v1/tools/color-palette`
+
+Accepts multipart form data with an image file. No settings field is needed.
+
+## Parameters
+
+This tool has no configurable parameters. Simply upload the image file.
+
+| Field | Type | Required | Description |
+|-------|------|----------|-------------|
+| file | file | Yes | The image to extract colors from |
+
+## Example Request
+
+```bash
+curl -X POST http://localhost:1349/api/v1/tools/color-palette \
+  -H "Authorization: Bearer si_your-api-key" \
+  -F "file=@photo.jpg"
+```
+
+## Example Response
+
+```json
+{
+  "filename": "photo.jpg",
+  "colors": [
+    "#304080",
+    "#e0a060",
+    "#f0f0f0",
+    "#203020",
+    "#a0c0e0",
+    "#806040"
+  ],
+  "count": 6
+}
+```
+
+## Response Fields
+
+| Field | Type | Description |
+|-------|------|-------------|
+| filename | string | Sanitized filename |
+| colors | array | Array of hex color strings, ordered by dominance (most frequent first) |
+| count | number | Number of colors extracted |
+
+## Notes
+
+- Returns up to 8 dominant colors, sorted by frequency (most common first).
+- The image is internally resized to 50x50 pixels for analysis, so the palette represents overall color distribution rather than small details.
+- Colors are quantized to the nearest multiple of 16 to reduce noise, then similar colors (within RGB Manhattan distance of 48) are merged to avoid near-duplicate entries.
+- The alpha channel is removed before analysis, so transparent areas are not considered.
+- This is a read-only endpoint. It does not produce a downloadable output file or a `jobId`.
+- HEIC, RAW, PSD, and SVG inputs are automatically decoded before analysis.