BugShot Privacy Policy

Effective date: July 14, 2026

BugShot (the "extension") values your privacy and collects and processes only the minimum information necessary. This policy transparently explains what information the extension handles.

1. Information We Collect

Platform Credentials and User Information

Assignees, CC (Watchers), and Mention Targets

When you use a field that selects a user — such as assignee, CC (watcher), or Slack mention — the extension queries the user directory of the connected platform (Jira, GitHub, Linear, Notion, GitLab, Asana, ClickUp, Slack) with the search term you enter and shows candidates (name, handle, avatar, email). To display already-selected users at the top of the list, the extension may additionally fetch those users' profiles (name, avatar, email). This lookup happens not only when you write an issue, but also when you set a default assignee in the Integrations tab. The identifiers of the targets you select are included in the body of the created issue (or Slack message) and sent to that platform.

Search terms and the candidate/profile lists returned are not stored on the device. However, the identifier and display name of the assignee, CC, or mention target you picked are stored on your device so they can be filled in again next time (as the default assignee in your integration settings, and as your last submission).

Page Data and Debug Information

When you run a full-page capture, the extension automatically scrolls the page from top to bottom until the capture finishes, and temporarily hides elements pinned to the screen (headers that follow you as you scroll, floating buttons, and so on), restoring them once the capture is done — this keeps the same header from being printed over and over in the image. While the capture runs, clicks and scrolling on the page are blocked so the result doesn't come out misaligned, and the page is scrolled back to where you were once it finishes. Because of this automatic scrolling, the page's own scroll-driven behavior (loading more content, the page's own analytics scripts, and so on) may run. That is behavior the page performs on its own; the extension does not collect or transmit anything beyond the captured image during this process.

When collecting network logs, sensitive headers such as authorization and cookie, and sensitive query parameters such as token and access_token, are masked automatically. Values of sensitive keys such as token, password, and secret in request/response bodies (JSON, form data) are also masked automatically. For real-time messages a page exchanges over WebSocket, only text frame payloads are collected (binary frames such as images or files are not collected), and the same body masking applies — though only to JSON-shaped frames; other text frames are collected verbatim. Console logs collect the messages a page prints verbatim (with no additional masking), so please be careful with debug capture on pages that print sensitive information to the console.

Console, network, and action logs may be collected not only from the current page but also from third-party frames embedded in that page (iframes — e.g., payment widgets, embedded SDKs), including clicks and input that happen inside those frames. This is because errors occurring in those frames may be needed to reproduce a bug; collected logs record the origin so you can distinguish and filter them by origin when attaching to an issue. The sensitive header/parameter/body masking above applies identically regardless of the frame's origin.

Element selection, style editing, and element capture also extend to elements inside cross-origin frames embedded directly in the page (iframes — payment widgets, embeds, etc.). When you select an element inside such a frame, the extension collects that element's selector, style information, and text, records the frame origin, and shows it in the issue's list of style changes. When you capture an element inside an iframe (by capturing the current tab's screen and cropping only that element's region), the frame's screen contents may be included in the screenshot. Frames nested inside another frame, or frames blocked by a security policy (sandbox), are not accessed internally.

User action logs record, in addition to the clicked element, input field, and navigation, reproduction steps for shortcut/special-key input (e.g., Enter, Esc, ⌘K — a shortcut entry carries no printable characters or field values), checkbox/radio toggles, dropdown selections, and drag actions (identifying information for the dragged element and the drop-target element). To make the entry readable, it also records the on-screen text of the element you clicked or dragged (its accessible name, up to 80 characters) — so it reads as "clicked Save" rather than a bare selector.

The values you type into input fields and pick from dropdowns are recorded verbatim (up to 500 characters) and attached to the issue, unless they are caught by the masking rules below. Knowing which value triggered the bug is what makes a report reproducible. Sensitive information is masked automatically (***) in two ways.

• By field type and label: type=password, autocomplete hints, and sensitive keywords found in the field's name, id, aria-label, associated label (a label element or aria-labelledby), or placeholder (password, card, cvv, ssn, token, and their Korean equivalents).
• By value shape: even when the label gives no clue, a value is masked if it looks like an email address or a run of 9 or more digits (phone, card, national ID, or bank account numbers).

The value-shape rule above applies not just to what you type, but to the element's on-screen text and field labels as well — if the name of the element you clicked looks like an email address or a long digit run, it is masked too.

In addition, content typed into rich-text editors (contenteditable — mail bodies, documents, message composers) is never recorded, neither as a value nor as an element name; only the fact that you typed is kept. Keystrokes while a sensitive field is focused are not recorded either. However, a value with no sensitive signal in either its label or its shape (a search term, ordinary text) is recorded verbatim, so on screens where you enter sensitive content, please turn off log attachment before submitting.

Console, network, and action logs are attached by default; you can turn each one off from the log cards on the drafting screen before submitting the issue.

While a capture is in progress (screenshot, report drafting, or video recording), navigating away (including to another site) does not interrupt the console/network/action logs — they are preserved, so an issue created that way may include debug logs from the pages visited during the capture.

When you reload a page on which you have previously started debug capture, resuming capture may retroactively include console/network/action logs from the early part of that page load (just before resuming). These early-load logs are held only temporarily in device memory until capture is restarted, and are not stored or transmitted.

App Settings

The LLM API key is stored obfuscated. You choose the provider yourself; presets ship base URLs for OpenAI, Anthropic, Gemini, Groq, OpenRouter, Together, and Ollama — when you use an AI feature, the material for the draft (title, body, style changes, annotated image, log summaries) is sent to the provider you picked. The action-log summary may carry unmasked input and selection values as-is (masked values are sent as ***). Nothing is sent if you don't turn the AI features on.

On browsers that support Chrome's built-in AI (Prompt API), drafts and CSS change suggestions can be generated with the on-device model without any external API call. In that case, data never leaves your device and no separate API key is required.

Anonymous Usage Analytics

The extension collects anonymous aggregate events to improve the product (effective June 19, 2026).

These events carry only the classification strings above and never include issue titles, bodies, URLs, or personally identifiable information. To distinguish the same installation, a random identifier (distinct_id) is generated once on install, stored on the device, and sent with subsequent events. This identifier is merely a random value and is not linked to any personal information such as email, account, or IP. To ensure the actual IP address is not stored, events are sent with the IP value set to 0.0.0.0, location estimation (GeoIP) is disabled ($geoip_disable), and personal profile creation is disabled ($process_person_profile: false). There is no separate opt-out (off) setting for this analytics.

Beyond the items above, the extension does not collect your browsing history, cookies, personally identifiable information, or the like.

2. Information Storage

All data is stored only inside your browser.

• chrome.storage.local: Jira, GitHub, Linear, Notion, GitLab, Asana, ClickUp, Slack integration settings (default project/repository/team and default assignee), your last submission (project, assignee, CC — including the identifier and display name of the person you picked), issue history, app settings, LLM provider settings (LLM API keys stored obfuscated)
• chrome.storage.session: Editing sessions (per tab, automatically deleted when the browser closes)
• IndexedDB: Video recordings, screenshot images, network logs, console logs, user action logs, inline editor images, user-attached files (local device only)
• Memory (temporary): 30-second replay frame buffer — not stored to disk; encoded to video and saved to IndexedDB only at the moment you perform a capture.

We do not store user data on external servers.

3. External Transmission

The extension transmits data only to the services below.

The OAuth proxy server only relays the token exchange and does not store or log user data. Linear and GitLab exchange tokens directly via PKCE without a proxy.

Local files you select yourself through the "file attachment" feature are, on issue (task) submission, uploaded as body attachments to each platform above (Jira, GitHub, Linear, Notion, GitLab, Asana, ClickUp), and to the message thread in the case of Slack. This feature is off by default and works only when enabled in settings.

When connecting to a GitLab self-managed instance with a PAT, the extension communicates directly with the instance address (an arbitrary origin) you enter. This access is covered by the required broad host permission (<all_urls>) granted at install and works without a separate permission dialog.

The LLM provider receives data only at the endpoint you configure yourself, and only when you explicitly run AI draft generation or AI styling. Access to that host is covered by the required broad host permission (<all_urls>).

When searching CC (watcher) mentions, the search term you enter is sent to each platform's user-search API, and the mention targets you select are sent as part of the issue body. All of this works only when you search and select yourself.

4. Third-Party Sharing

We do not sell, share, or transfer the information we collect to third parties. Data is transmitted to a given platform, or to the LLM provider you configure, only when you create an issue or request an AI draft or AI styling yourself.

5. Data Deletion

• Removing the extension: chrome.storage data is deleted automatically.
• Media / log data: You can clear site data in your browser settings, or delete items individually from the issue list inside the extension.
• Disconnecting a platform: Disconnecting on each platform's (Jira, GitHub, Linear, Notion, GitLab, Asana, ClickUp, Slack) integration tab deletes the stored credentials.
• Deleting an LLM provider: Disconnecting the provider in settings deletes the stored settings.

6. Permissions Notice

Extension Permissions

Host Permissions

The extension has a single host permission: <all_urls>.

• <all_urls> (all sites) — the required permission for performing DOM selection, screen capture (captureVisibleTab), and console/network log collection on arbitrary web pages, and for communicating with the LLM provider / GitLab self-managed instance you configure, as well as the API servers of the issue trackers / Slack you connect and the OAuth proxy. Granted at install and shown on the install screen as "Read and change your data on all sites."

There is no separate per-platform host permission; all of the communication above happens under the <all_urls> permission. For the external destinations to which data is actually transmitted (each platform's API server, the OAuth proxy, etc.), see "3. External Transmission."

Where the Broad Host Permission (<all_urls>) Is Used

<all_urls> (all sites) is a required permission granted at install (shown on the install screen as "Read and change your data on all sites"). It is needed for the extension's core features that operate on arbitrary web pages, and there is no separate runtime permission dialog. Main uses:

• DOM selection / style editing: Picking an element on any web page to collect information and preview styles
• Screen / full-page capture and 30-second replay: Current-tab screen capture (captureVisibleTab) does not work with ordinary host permissions and requires <all_urls>. Full-page capture calls the same API repeatedly while scrolling the page, stitching in the areas that were off-screen. Capture and log collection continue even when you navigate to another site, without the side panel closing.
• Console / network log collection: Recording logs on arbitrary pages (and iframes)
• AI draft / AI styling: Transmitting to the LLM provider endpoint you configure yourself (when you explicitly run an AI draft or AI styling)
• GitLab self-managed: PAT communication with the instance (an arbitrary origin) you enter
• Style value enrichment: To accurately display the "author-specified" styles of a selected element, the extension reads, in the background and without credentials (credentials:omit), the external stylesheets (cross-origin CSS files) the page references. Only public http(s) hosts are targeted (loopback, internal networks, and private IPs are blocked), and the CSS received is used only on the device and not transmitted to third parties.

Each feature transmits data only when you turn it on or run it yourself. The permission itself is granted at all times from install, but you can narrow the access scope in Chrome settings (Extensions > BugShot > Site access).

7. Changes

If this policy changes, we will provide notice through this page.

8. Contact

Privacy inquiries: ox501501@gmail.com