Add privacy tools, inline file diff, and responsive layouts

.github/workflows/ci.yml

@@ -15,7 +15,7 @@ jobs:
           node-version: 24
           cache: npm
       - run: npm ci
-      - run: npm audit --omit=dev --audit-level=high
+      - run: npm run security-check
       - run: npm run privacy-check
       - run: npm run lint
       - run: npm run typecheck

AGENTS.md

@@ -0,0 +1,9 @@
+<!-- BEGIN:nextjs-agent-rules -->
+
+# This is NOT the Next.js you know
+
+This version has breaking changes — APIs, conventions, and file structure may all differ from your training data. Read the relevant guide in `node_modules/next/dist/docs/` (resolved from this file's directory; in monorepos the `next` package may not be visible from the repo root) before writing any code. Heed deprecation notices.
+
+This block is written and re-added by `next dev` — verify at `node_modules/next/dist/server/lib/generate-agent-files.js`. Removing it from a diff only re-creates the uncommitted change; committing it with your work keeps the tree clean.
+
+<!-- END:nextjs-agent-rules -->

README.md

@@ -24,11 +24,34 @@ ToolStack is an open-source collection of utilities designed around a strict loc
 ### Privacy & Security
 Image Metadata Remover, Data Anonymizer, File Hash & Checksum, Tracking URL Cleaner, Password & Passphrase Generator, JWT Inspector, Browser Privacy Inspector, Text Privacy Cleaner, AES-GCM File Encryption, File Type & Signature Inspector, Content Credentials (C2PA) Inspector, and HMAC Generator & Verifier.
 
+### Private sharing workflows
+- Support File Sanitizer: inspect outgoing logs, JSON and HAR files with batch-wide aliases for supported identifiers. HAR bodies are excluded by default, credentials are redacted, and original filenames and alias maps are omitted from downloads. Processing runs in a time-limited worker.
+- Screenshot Redactor: draw opaque redaction areas or enter pixel coordinates, then export a freshly rendered PNG. Manual review is required; this version does not include OCR or automatic detection.
+- Document Privacy Inspector: inspect DOCX, XLSX and PPTX XML parts for document properties, comments, tracked changes, hidden-content markers and external references. Clear supported document properties and reinspect the generated file. Document content, comments and tracked changes are not removed. Encrypted, legacy and ZIP64 documents are unsupported; compressed and expanded size limits apply.
+- Data Anonymizer now also masks linked CSV / JSON datasets with per-column actions and shared alias groups. It can remove columns, mask values, shift ISO dates and group numbers into bands. This is pseudonymization, not a guarantee of anonymity. Alias values are text labels and CSV exports escape formula-like values.
+- Calendar Privacy Exporter: rebuild a selected date window as individual Busy events, using new IDs and excluding original meeting details. Recurrence, individual exceptions and embedded time zones are supported. Unsupported recurrence forms, unresolved named zones and processing limits stop the export. Busy times are still disclosed; the snapshot does not update automatically.
+
+### Appearance and navigation
+A sun/moon toggle switches between light and dark appearance across all tool pages and native form controls. The site follows the system appearance until a mode is chosen. Only the appearance preference (`toolstack-theme`) is saved in local storage; working input and generated output are not saved there. The directory includes search, category counts, compact cards and grouped browsing.
+
+### Text & File Diff
+- The existing `/text-diff` tool now has two editable CodeMirror panes, with changes highlighted inside the inputs, aligned blocks, individual-change transfer, undo/redo, next/previous navigation, wrapping, whitespace markers and folding of unchanged sections. Source highlights and merges always use exact text. Very expensive editor diffs may use a visibly disclosed approximate highlight; the separate worker review requires an exact completed comparison.
+- Private Review compares text, JSON, headered CSV or literal single-line `.env` files. JSON object key order and formatting are excluded; CSV columns are matched by header. Match arrays of records with one or more exact key columns; missing/duplicate identities stop comparison. Common record reordering is counted separately using a longest increasing subsequence. Exact unique moved text blocks are annotated when they appear as distinct additions/removals.
+- Explicit comparison rules support ASCII whitespace, string case, line-ending style, ISO timestamp/UUID-shaped text, literal text, structured field paths and JSON numeric tolerances. Excluded field/line differences remain inspectable. Expected field paths flag unignored changes outside the selected paths. Patterns use JSON Pointer-style paths (`~0` / `~1` escaping), `*` for one level and `**` for any depth; this is an explicit review aid, not a program-behavior validator.
+- Comparison is performed on original values before masking. Per-comparison aliases preserve equality and differences between detected credentials, common identifiers and manually supplied values. No alias map is exported. Detection is best-effort; paths, keys, free text and context require review. Private Review is read-only and never overwrites source documents. File details reveal newline/BOM differences, hidden Unicode and an illustrative list of common lookalikes.
+- Select individual changes, optionally include up to five context lines, preview outgoing text, then download a standalone HTML or text report. HTML has no script/external assets and escapes input markup; only selected redacted excerpts and comparison-rule descriptions are serialized. Reports exclude original filenames and full source documents. Original source downloads and exact `.diff` patches are separate, explicitly unmasked actions.
+- Inputs stay in tab memory. Up to 1 MB and 20,000 line endings per side; UTF-8 and UTF-16 with BOM are decoded strictly. Source downloads retain imported encoding and existing newline characters; new typed lines use LF. Structured limits, duplicate-key detection and safe-integer checks prevent ambiguous comparisons. Worker timeout is eight seconds; bounded diff failure produces an error rather than partial results. Binary files, formatted Office/PDF comparison and shell evaluation are unsupported.
+
 ### Images
 Image Compressor, Image Resizer, Image Cropper, Image Format Converter, and Placeholder Image Generator.
 
 ### Developer / Data
-JSON Toolbox (including structural diff), Regex Tester & Explainer, Easy Regex Generator, Text Diff Checker, sanitized Markdown Editor, CSV to JSON/XML Converter, XML Toolbox, YAML ↔ JSON Converter, AST-based CSS Specificity Calculator, CSP Builder & Analyzer, Cron Expression Explainer, HTTP Status Code Explainer, HTML Tag Explainer, Encoder / Decoder, and UUID / ULID Generator.
+JSON Toolbox (including structural diff, TypeScript and JSON Schema inference), Regex Tester & Explainer, Easy Regex Generator, Text Diff Checker, sanitized Markdown Editor, CSV to JSON/XML Converter, XML Toolbox, YAML ↔ JSON Converter, AST-based CSS Specificity Calculator, CSP Builder & Analyzer, Cron Expression Explainer, HTTP Status Code Explainer, HTML Tag Explainer, Encoder / Decoder, and UUID / ULID Generator.
+
+- JSON inference merges fields across array samples, marks missing fields optional and retains mixed types. Schemas describe the sample rather than a complete API contract. Sample values are not included in generated definitions.
+- Support File Sanitizer includes a POSIX cURL-to-JavaScript converter. It parses text without executing commands or sending requests. Sharing mode replaces query values, JSON body values and most header values; URL paths, field names and content-negotiation headers remain visible. Browser-controlled headers are omitted with notices; file references, shell expansion and unsupported flags stop conversion. Output always omits browser credentials.
+- Synthetic Dataset Generator creates seeded fixtures with unique per-table IDs and valid references between tables. Customize types, ranges, choices and row counts; preview tables or export CSV, JSON and a ZIP with the schema. Test names are labels and generated emails use `example.invalid`. CSV formula-like text is escaped. Seeded UUIDs are fixture data, not cryptographic identifiers. Limits: 8 tables, 20,000 total rows and 300,000 cells; processing runs in a time-limited worker.
+- QR Toolkit generates text/URL, personal Wi-Fi and vCard contact codes with bundled `qrcode`. PNG/SVG exports contain the encoded values, including Wi-Fi passwords. Read or paste PNG/JPEG/WebP images with bundled `jsqr` in a time-limited worker; targets are displayed as inert text, never opened or checked over the network. Camera access is not required. Images are limited to 20 MB / 20 million pixels and scaled to at most 1,600 pixels per side for decoding; small or complex codes may need a sharper crop.
 
 ### Text, Accessibility, Converters & Generators
 Text Statistics / Analyzer, Find & Replace, List Cleaner & Formatter, Text Contrast Checker, Color Code Converter, Unit Converter, Timestamp & Date Toolkit, and Lorem Ipsum Generator.
@@ -37,7 +60,7 @@ Text Statistics / Analyzer, Find & Replace, List Cleaner & Formatter, Text Contr
 
 ToolStack registers a same-origin service worker in production. The home page, offline fallback and ToolStack's own static assets are cached, and successfully visited tool pages can be reopened from the cache when offline.
 
-User-selected files, generated files and tool input are **not** deliberately placed into the service-worker cache.
+User-selected files, generated files and tool input are **not** deliberately placed into the service-worker cache. Bundled PDF.js and qpdf runtime assets are cached after successful use, alongside application chunks.
 
 ## Development
 

eslint.config.mjs

@@ -10,6 +10,8 @@ export default defineConfig([
     'out/**',
     'build/**',
     'coverage/**',
+    'public/qpdf/**',
+    'public/pdfjs/**',
     'next-env.d.ts',
   ]),
 ]);