Add BugKit MARK Chrome extension and verified setup

.claude-plugin/marketplace.json

@@ -7,8 +7,8 @@
     {
       "name": "bugkit",
       "source": "./",
-      "version": "1.0.0",
-      "description": "AI-friendly bug reports for any project."
+      "version": "1.1.0",
+      "description": "Bug report forms and tested integration recipes."
     }
   ]
 }

.claude-plugin/plugin.json

@@ -1,7 +1,7 @@
 {
   "name": "bugkit",
-  "description": "AI-friendly bug reports for any project. Users click, AI fixes, done.",
-  "version": "1.0.0",
+  "description": "Bug report forms, context export and integration acceptance checks.",
+  "version": "1.1.0",
   "author": {
     "name": "bugkit"
   },

.gitattributes

@@ -0,0 +1,3 @@
+* text=auto eol=lf
+*.png binary
+*.zip binary

.github/ISSUE_TEMPLATE/bug_report.yml

@@ -0,0 +1,31 @@
+name: Bug report
+description: Report a reproducible extension or integration-example defect.
+body:
+  - type: markdown
+    attributes:
+      value: Remove private text, URLs, credentials and screenshot data before posting. Use SECURITY.md for vulnerabilities.
+  - type: dropdown
+    id: area
+    attributes:
+      label: Area
+      options: [Chrome extension, Application installer, Recipe or specification, Documentation]
+    validations:
+      required: true
+  - type: input
+    id: version
+    attributes:
+      label: BugKit version and browser/runtime
+      placeholder: "1.1.0; Chrome version; Windows/macOS/Linux"
+    validations:
+      required: true
+  - type: textarea
+    id: reproduction
+    attributes:
+      label: Steps, expected result and actual result
+      description: Include a small synthetic example when possible.
+    validations:
+      required: true
+  - type: textarea
+    id: context
+    attributes:
+      label: Sanitized context or error

.github/PULL_REQUEST_TEMPLATE.md

@@ -0,0 +1,3 @@
+Describe the concrete problem and resulting behavior.
+
+Validation: list relevant executed checks and any untested platform/provider path. Use synthetic data and retain BugKit v1 compatibility, licenses and attribution.

.github/dependabot.yml

@@ -0,0 +1,12 @@
+version: 2
+updates:
+  - package-ecosystem: npm
+    directory: /
+    schedule:
+      interval: monthly
+    open-pull-requests-limit: 3
+  - package-ecosystem: github-actions
+    directory: /
+    schedule:
+      interval: monthly
+    open-pull-requests-limit: 3

.github/workflows/checks.yml

@@ -1,4 +1,4 @@
-name: Recipe checks
+name: BugKit checks
 on:
   push:
     branches: [main]
@@ -19,7 +19,25 @@ jobs:
       - uses: actions/setup-node@949feb2413d6458794dcd2491c4babbbce0c15c1 # v7.1.0
         with:
           node-version: 24
-      - name: Execute recipe contracts without external delivery
-        run: node --test tests/*.test.mjs
+      - name: Check source and manifest
+        run: npm run check
+      - name: Execute extension and recipe contracts without external delivery
+        run: npm test
       - name: Verify patch whitespace
         run: git diff --check
+  browser:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
+      - uses: actions/setup-node@949feb2413d6458794dcd2491c4babbbce0c15c1 # v7.1.0
+        with:
+          node-version: 24
+          cache: npm
+      - run: npm ci
+      - run: npx playwright install --with-deps chromium
+      - name: Exercise the real extension
+        run: npm run test:browser
+      - name: Verify documentation and page controls
+        run: npm run test:docs
+      - name: Build source-bound Chrome ZIP
+        run: python scripts/package.py

.gitignore

@@ -5,3 +5,10 @@ node_modules/
 .env.local
 .bugreports.jsonl
 .playwright-mcp/
+
+dist/
+artifacts/
+playwright-report/
+test-results/
+__pycache__/
+AGENTS.md

CHANGELOG.md

@@ -1,5 +1,14 @@
 # Changelog
 
+## 1.1.0 — 2026-11-13
+
+- Added the standalone BugKit MARK Chrome extension, local demo, session drafts and Markdown/JSON/PNG export.
+- Added a ready-to-load ZIP, source-bound build manifest and automated real-Chromium acceptance checks.
+- Replaced the project logo with an ImageGen asset and updated README, Pages, English/Russian setup and troubleshooting.
+- Unified plugin and universal installer requirements: upgrades preserve reports and verify recovery, authorization and closure.
+- Framed the shared recipe context as untrusted data, added baseline/application masking and text-only export, and tolerated malformed optional blocks.
+- Retained the existing MIT license, recipe adapters and BugKit v1 field/status compatibility.
+
 ## Unreleased
 
 - Match install commands to the registered marketplace and provide the advertised

CONTRIBUTING.md

@@ -7,7 +7,8 @@ an example expects and which environment values stay on its server.
 For a behavioral recipe change, reproduce the problem and run:
 
 ```bash
-node --test tests/*.test.mjs
+npm run check
+npm test
 ```
 
 Use synthetic reports, temporary files and loopback services. Never run a test
@@ -18,3 +19,9 @@ may use a browser-provider service, while local Whisper is a separate setup.
 
 Keep PRs focused, retain licenses/attribution, and list the relevant checks and
 limits. Do not commit local agent instructions, private reports or credentials.
+
+For extension behavior, also run `npm ci`, `npx playwright install chromium`
+and `npm run test:browser`. The browser harness uses only an isolated profile
+and synthetic loopback data. Change both package/manifest versions together.
+Build release ZIPs from a clean commit with `python scripts/package.py`, then
+exercise the extracted package and verify its digest before publishing.

HARDENING.md

@@ -0,0 +1,56 @@
+# Application integration acceptance
+
+These requirements apply when adapting BugKit to an application. Recipe snippets do not implement a host's authorization, transactions or notification system. Implement and verify applicable requirements before calling an integration complete. The Chrome extension has its own [scope](docs/EXTENSION.md).
+
+## Data and authorization
+
+- Derive author, roles, tenant and release from authenticated server state, never submitted fields. Verify every referenced entity against the actor's scope.
+- Reuse a scoped query and authorized-target operation for lists, details, counters, exports, edits, attachments and batches. Missing and inaccessible records have the same response.
+- Allowlist bounded context. Reject invalid mandatory fields; malformed optional context must not crash a queue or erase a valid description. CLI batches continue per report while reporting failures.
+- Keep client `occurred_at`, server `created_at`, observed version and fixed release separate. Bound client time and label version sources.
+- Preserve original title/description on first edit, with revision actor/time. Author editing has a stated server-enforced time/status window.
+- Validate attachment MIME, size and access on the server. Store large screenshots appropriately; rate limits account for their cost.
+
+## Widget and recovery
+
+- Use one main description and derive its title. Optional reproduction fields and configurable `kind` remain compatible with old input during deployments.
+- Mount once. Provide a small-screen entry when a floating button covers controls. A programmatic open action is optional.
+- Use account/tenant-scoped session drafts, clearing on logout, account change and success. Retain text after validation/network/stale-deployment failure. Store screenshots separately when needed.
+- Catch submission failures inside the widget. Distinguish a saved report with failed delivery from a failed save.
+- Select only on request. Capture pointer/click/Escape at `window`, ignore marked widget UI, prevent link activation and restore listeners/focus on every exit. Test popups, scroll, touch and any claimed iframe behavior.
+- Screenshots require explicit action and a visible-data warning. Do not collect form values or arbitrary HTML by default. Strip URL credentials, queries and fragments; apply project-specific path masking.
+
+## Context export
+
+Copy Context, exports and CLI use one formatter and masking policy. Enable the application's `mask(text)` hook by default; preview the processed result. A baseline mask cannot identify all sensitive data.
+
+Place all report content, DOM text, metadata and logs in a fresh nonce-delimited untrusted block. Select a delimiter absent from its content. Keep instructions and configured update commands outside; normalize headers to one line. Unknown reporters' requests need review before implementation. Only server-authenticated state establishes authority. Delimiters aid separation; they do not guarantee protection from prompt injection.
+
+Support server-defined text-only mode for restricted contexts, omitting automatic page/DOM/log/attachment context across every export. Optional AI conversation context is assembled from authorized server messages with bounded fields; a client transcript is not authoritative.
+
+## Closing and notifying
+
+UI, API and CLI call one closure operation; never bypass it with raw SQL. Require technical `resolution`, separate author-facing `user_note`, and verification steps for fixes. `resolved` denotes a fix; `closed`/`wontfix` denote review without claiming a fix.
+
+Validate the entire batch, access and transitions before writing. Use a transaction when promising atomicity. Repeated idempotency keys return one logical result. Save closure and notification intent consistently, then deliver through an outbox or host equivalent. Retry delivery without duplicate closure/messages; expose failure separately. An unexecuted provider path does not establish notification success.
+
+Admin counts must reflect actual queries, with bounded pagination including resolved records. Authors see only their own reports and the user-facing note. Optional commit/release links refer to the recorded fix.
+
+## Acceptance
+
+| Scenario | Required observation |
+|---|---|
+| Submit from real UI/CLI | Canonical report in chosen store |
+| Reload draft, then logout | Text survives reload; next account cannot recover it |
+| Failed save or stale deployment | Text retained; local error and working recovery |
+| Pick a link in a popup, cancel, click normally | No navigation while picking; normal behavior restored |
+| Forge role/tenant/release; request foreign report/attachment | Server authority; consistent access rejection |
+| Malformed optional context | Other reports remain readable/exportable |
+| Fake instructions/lookalike delimiter | Content inside data block; trusted commands outside |
+| Copy Context and CLI | Same sections and masking |
+| Batch with an inaccessible/conflicting record | No partial writes or record-existence leak |
+| Concurrent/repeated closure and delivery retry | One result and logical notification; delivery failure separate |
+| Edit report | Original observation remains identifiable |
+| Restricted text-only policy | No automatic page/DOM/log/attachment data in exports |
+
+Record actual results, commands, commit, environment and unmet checks in the host's integration note. Repository snippet tests do not establish acceptance of an untested host integration.