Context Pack Help
Capture a page or selection, review it, then copy Markdown without building a pack. To combine web captures and local files, keep up to 10 retained sources in the workspace and export one editable context-pack.md.
Install Context Pack with Add to Chrome, then choose the recipe for your source. These steps use the current extension, not an online upload form.
Open the page and let the content you need load. Open Context Pack and choose Capture page. Check the editable Review capture result, then choose Copy Markdown without adding a source or building a pack. For selected text, highlight it first and choose Capture selection. See Quick Copy; use the pack recipe when you want a combined file.
Open the extension and choose Open current pack. In the workspace, use Choose files or drag and drop a supported local file. After validation and conversion, it appears in the source list without a browser-capture review panel. Build the pack and inspect the Markdown before exporting. Check PDF and file limits.
Choose Add to pack after reviewing each browser capture; imported local files go straight into the source list. Keep the workspace open to collect another source. Reorder or exclude sources, then choose Build context pack and inspect the output. Finish with Copy Markdown or Download .md. Up to 10 retained sources share one pack; closing or reloading discards the editable session.
Choose the action for your source, or see actual table-to-Markdown output.
Choose the capture action, review the result, then copy it or add it to your pack. Read the release notes.
| Your source | Action | Output and next step |
|---|---|---|
| Web page or selected text | Capture page / Capture selection | Editable review. Copy Markdown now, or Add to pack to combine sources. |
| One visible HTML table | Capture table / Capture selected table | A table draft with layout warnings when needed. Copy Markdown or Add to pack. |
| Visible page or selection links | Extract links | Filter a link list; Copy URLs, Copy Markdown or Add to pack. Destination pages are not fetched. |
| Selectable-text PDF | Choose files or drop a local file | Converted text becomes a source. Build and inspect the pack before exporting; see PDF limits. |
| DOCX document | Choose files or drop a local file | Converted text becomes a source. Build, review, then copy or download the Markdown. |
| Markdown or UTF-8 text | Choose files or drop a local file | Add your text to the source list, then build and export a combined pack. |
Link labels that look like Markdown stay readable as literal labels in Preview, not extra nested links. Code and ordinary formatting remain intact; Copy Markdown still exports your unchanged draft.
Choose Capture page or Capture selection. Review and edit Markdown, switch to read-only Preview, then choose Copy Markdown. Copy uses the current draft and leaves review open; nothing is added or built. Add to pack is optional. Cancel discards the pending review without changing earlier sources or output. Direct-PDF review is unchanged.
On a regular webpage, choose Capture table in the popup or workspace. With multiple tables, use Choose a table to select its caption or numbered title; one table opens directly in review. Review Markdown or Preview, then Copy Markdown or Add to pack for one Table source. Switching tables keeps separate edited drafts until Add or Cancel; capture again to refresh the page snapshot.
Already looking at the table you want? Highlight any ordinary text in a cell of that one main-page HTML table, then right-click the highlight and choose Context Pack → Capture selected table. No special word is required; changing a dropdown option is not page-text selection. Review shows the whole sanitized outer table, not just highlighted cells, without a redundant chooser. To keep only the highlighted text, use Capture selected text.
Want to browse tables instead? Open the extension and choose Capture table. Discovery shows its chooser only when multiple tables are found. Review/edit one table, then Copy Markdown or Add to pack. Capture table again returns to discovery.
If capture is unavailable: outside-table or multiple-table selections are rejected with guidance. Select ordinary cell text in just one table and retry. Hidden/editable regions and embedded-frame tables are not captured. If the right-click action is missing inside an embedded frame, open that frame's page in its own tab to retry. Your pending review and existing pack stay unchanged.
Unambiguous merged cells repeat their content with a warning. Complex, nested or headerless tables, and inline code with a backslash immediately before a pipe, fall back to explicit row/cell text when a grid would misrepresent them. Images, including their alt text and image-only links, are omitted from table capture; they are not downloaded. Review Markdown and removal warnings before sharing. Load the rows you need before capture; this action captures loaded HTML tables, not PDF tables or virtualized grids.
Highlight a visible main-frame region for selected links, or clear the selection for visible page links, including navigation/footer links. Choose Extract links in the popup or workspace. Set Text or URL contains, All/Internal/External destinations and Remove duplicate URLs, then Apply filters. Internal means the same protocol, host and port. Replacing an edited draft requires confirmation.
Copy URLs reads the current edited Markdown and exports unique sanitized HTTP/HTTPS destinations in document order, excluding images, code and raw HTML. Copy Markdown exports the draft exactly. Add to pack creates one Links source; Copy, filters and Cancel leave the existing pack unchanged. Extracting links does not open, crawl or fetch their destinations.
Discovery rejects more than 20 eligible tables; direct selected-table capture targets one even when the page contains more. Both enforce DOM/range safety limits and 500,000 combined capture UTF-16 code units. Link extraction rejects more than 1,000 inspected anchors. No silent truncation. Empty or oversized drafts cannot be copied or added. The 10-source and retained-pack limits apply to Add, not standalone Copy.
For clipboard denial, return to Markdown and select/copy manually. Copy URLs stays disabled while parsing, when no usable URLs remain, or after a bounded-parser failure; edit/retry or use Copy Markdown. See capture troubleshooting.
Capture is main-frame only and excludes hidden, editable and form-input content. Visible labels and retained link query parameters may still be private: no automatic PII redaction. Preview/filtering/table/link capture make no source requests; the existing direct HTTPS PDF exception is unchanged. Copied text can outlive the workspace in your OS clipboard.
The same neutral Regions fixture used in the table screenshot, converted with Context Pack 1.3.0 without editing the Markdown. The numbers are sample data, not product usage figures.
To try it here, highlight ordinary cell text in the table below, then right-click Context Pack → Capture selected table. See the table capture steps and release status.
| Region | Metric | Value |
|---|---|---|
| Europe | Users | 120 |
| Teams | 8 | |
| Asia | Not published | |
Markdown cannot merge cells. Review warns that merged cells were expanded: repeated values below represent the same original cell, not additional observations.
Regions
| Region | Metric | Value |
| --- | --- | --- |
| Europe | Users | 120 |
| Europe | Teams | 8 |
| Asia | Not published | Not published |Tables that cannot be represented safely use explicit row/cell text instead of a misleading grid. Read the table capture steps and warnings.
What you see first
The first run opens Bring several sources into one workspace. Choose Start adding sources or skip the tour. After that, the left column adds and manages sources; the right column holds the built Markdown output.
Only the “tour completed” preference is remembered. Use Help in the workspace header to open the tour again. Closing or reloading the workspace discards its editable session.
On a regular page, open Context Pack and choose Capture page. The popup closes; Chrome opens or focuses the Context Pack workspace and stages a Review capture panel. Check the title, sanitized URL, warnings, and editable Markdown, then choose Add to pack.
The popup closes and Chrome opens or focuses the dedicated workspace for review.
See the flow → 02 · TabsCapture another tab without losing the sources already in the pack.
See the flow → 03 · FilesPDF, DOCX, Markdown, and text files convert locally into the source list.
See the flow → 04 · OutputArrange sources, build once, edit the output, then copy or download.
See the flow →Leave the workspace open, move to another page, then capture again from the toolbar, right-click menu, or shortcut. Chrome focuses the existing workspace for review. After Add to pack, the first source remains and the counter becomes 2 included · 2 / 10.
Open the workspace, then choose or drop selectable-text PDF, DOCX, Markdown, or UTF-8 text files. They convert locally and enter the source list after validation. Unlike browser captures, local files do not use the staged Review capture editor.
Review document tables, lists and complete headings before sharing. Read the maintenance note.
Simple rectangular DOCX tables keep their rows and columns, including empty cells. Ambiguous merged or uneven rows and richer cells use warned row/cell text. Cell numbers mean source order, not reconstructed visual columns; compare the result with the original document.
Nested DOCX lists retain their hierarchy. A document heading with line breaks becomes one complete heading, with a flattening warning; check it in Sections and Preview. Ordinary paragraph breaks remain breaks.
Read the source warnings and Markdown, choose Build context pack, then inspect Preview. Return to Markdown to edit; Copy Markdown and Download .md export that editable text. Document troubleshooting.
Use the left column to include, exclude, reorder, or remove retained sources. Choose Build context pack; the right column shows editable Markdown and a local token estimate. Choose Preview for a read-only rendered view; return to Markdown to edit. Finish with Copy Markdown or Download .md.
Preview formats your current output locally. Raw HTML remains text and images become alt-text placeholders; no images or remote resources load. Copy Markdown and Download .md always use the editable Markdown, not the rendered view.
Supported visible actions, selected non-editable choices, disabled and pressed/expanded states retain meaning. Input values and hidden choices stay excluded. Labels and choice summaries are capped at 200 Unicode code points with a warning when simplified. Complex tables use explicit row/cell text and span annotations when a grid would mislead. This is not a visual replica or automatic PII redaction.
Large or complex output may exceed the bounded renderer. Return to Markdown to edit, copy or download. Changing sources clears stale output; build again before previewing or exporting. Closing or reloading still discards the editable session.
Our sample page becomes one source in context-pack.md, keeping its headings, numbered list and table. Add other pages or supported files to the same workspace to build a combined pack.
Captured with Context Pack 1.2.0 and built without editing the Markdown. This is one reproducible webpage example, not a guarantee for every website, PDF or DOCX.
Open the sample page in a new tab. Choose Capture page, review the result, then Add to pack and Build context pack.
The capture review reports that a broader page view was used to retain more structured content. Review that warning before adding. Scroll inside the example to see the complete output.
# Context Pack
## Source S1 — Local Context Pack Field Guide
- Type: Web Page
- URL: <https://tryaiblock.com/tools/context-pack/demo/>
- Sections: A practical guide to local context packs.; Keep sources inspectable; Review before reuse; Match the source to the job
## A practical guide to local context packs.
Bring focused source material together, keep every boundary visible, and inspect the Markdown before it leaves your browser.
---
## Keep sources inspectable
A useful context pack preserves source order and provenance. Start with a small set of relevant documents, remove repeated navigation and visual noise, then review the converted text before combining it.
1. Choose only the sources needed for the task.
2. Review titles, links, warnings, and document boundaries.
3. Build the pack and inspect the complete Markdown.
## Review before reuse
Review each source before combining it. A visible source order makes the final context easier to inspect, edit, and explain before export.
## Match the source to the job
| Source | Useful for |
| --- | --- |
| Readable page | Focused articles and documentation |
| PDF or DOCX | Reference material and working drafts |
| Markdown or text | Notes, briefs, and structured prompts |
Review the actual content and warnings before sharing. This file does not preserve the editable workspace; copy or download your Markdown before closing it.
Toolbar, right-click, and keyboard actions all stage browser content in the current workspace before it is retained.
Capture page reads the current page. Highlight text first to use Capture selection. Open current pack focuses the workspace without capturing.
Use Context Pack → Capture this page or, after highlighting text, Capture selected text.
Alt+Shift+M on Windows/Linux or Option+Shift+M on macOS captures the current page. Inspect or change the assigned command at chrome://extensions/.
New pack asks for confirmation when work exists. Closing or reloading always clears the editable session.
The limits keep local conversion and review predictable and prevent partial content from entering a pack.
On an eligible direct HTTPS URL whose path ends in .pdf, the popup changes to Capture PDF. Context Pack makes the requested fetch, validates the response, and parses it locally.
Embedded, custom, blob, protected, or plain HTTP PDF viewers are not direct-PDF capture targets. Download the file and add it locally, or open a supported direct HTTPS .pdf URL.
Markdown/text: 1.5 MB each. PDF/DOCX: 20 MB each and best under 10 MB. PDF: 30 pages per file.
10 retained sources, 50 MB total retained source bytes, 60 PDF pages, and 500,000 generated UTF-16 code units in the final pack.
No OCR for scanned/image-only PDFs, encrypted PDF or DOCX, legacy .doc, or malformed non-UTF-8 Markdown/text files.
A rejected capture or file shows a clear error and does not add partial content or replace sources already retained.
Context Pack does not passively read tabs and does not require a TryAIBlock account.
Chrome grants temporary access only after you invoke capture on the current tab.
Page, selection, and file conversion runs in the extension. An eligible direct PDF makes one requested HTTPS fetch from its current URL.
The source list and editable output stay in the loaded workspace. Copy or download before closing if you need to keep the result.
Chrome-internal pages such as chrome://version/ cannot be captured. Open a regular HTTP or HTTPS page, invoke Context Pack there, and choose Capture page or Capture selection. If source access expired or its tab closed, return to or reopen that source first. Keep the workspace open: do not reload it or choose New pack to recover capture. See capture error messages and recovery steps.
Capture converts readable content from the current main page; it is not a screenshot or a crawl of linked pages. Hidden or unloaded content, media, and unsupported embedded frames may be omitted. Let the desired section load, capture again, and inspect the warnings before adding. Use Capture selection for a specific visible passage. Do not expect interactive dashboards to retain every visual or control.
No OCR is included: PDFs need selectable text. DOCX conversion preserves supported document structure, not page layout; embedded images are omitted and complex or merged tables may lose fidelity. Review the output before use.
No. It does not bypass access or download restrictions. For a custom or embedded viewer, use a supported direct HTTPS PDF URL or add a local copy only when the site allows it. Check capture troubleshooting.
That is the intended workspace-reuse flow. It keeps earlier sources together and avoids duplicate Context Pack workspaces.
Highlight visible text in the main page first, then invoke Capture selection. Content inside unsupported embedded frames is not part of the main-page selection.
A browser or site viewer may be a webpage rather than a supported direct HTTPS PDF response. Download the PDF and add it locally when the site allows it.
It is not silently persisted. It remains only in the loaded workspace memory until you close or reload it; copy or download the generated Markdown to keep it.