You updated uBlock Origin. The planted filter is still there. I built a 30-second check for it.
On 15 November 2026, researcher Smitty van Bodegom published a one-click universal XSS in uBlock Origin (Hacker News #6, 607 points / 177 comments, checked 17 November 2026). A malicious page could drive the element picker with postMessage plus one clickjacked click and smuggle a newline into the picker's opt field to append a full trusted scriptlet filter - for example trusted-rpnt scoped to google.com. Picker-created rules are saved to My filters with trustedSource: true, so the scriptlet runs JavaScript on every site the victim later visits. GitHub advisory GHSA-cmqw-9r2w-6wv2 rates it High; versions 1.75.2 and lower are vulnerable, 1.75.4 is patched, and 1.76.4 or newer (as of 17 November 2026) is the version to be on. No exploitation in the wild has been reported.
Updating closes the hole. It does not remove a filter the attack already planted. I verified this myself: I downloaded the official uBlock Origin builds 1.75.2 and 1.76.4 and compared the function that saves picker rules - appendUserFilters in js/storage.js is byte-identical, 3,137 characters, in both. No release in the fix series deletes a planted filter (Medium / Chinmay Shringi, 16 November 2026). So the correct action after updating is a 30-second check of My filters - and nobody runs that check, because no tool does it. I built filterguard to be that tool, plus an audit of every installed extension by permission and host-scope risk, in one local command with a CI gate.
What it does
filterguard is a zero-dependency TypeScript CLI (Node 20+). Everything runs locally; no data leaves the machine.
filterguard scan [--profile-dir <dir>] [--browser chrome|chromium|brave|edge|firefox] [--json] [--save-baseline <file>]
filterguard filters [--filters <file>] [--json] [--save-baseline <file>]
filterguard check [--profile-dir <dir>] [--filters <file>] [--baseline <file>] [--json]
filterguard diff <baseline-a> <baseline-b> [--json]
filterguard explain <rule-or-extension-id> [--profile-dir <dir>] [--filters <file>] [--json]
filterguard init
-
scandiscovers browser profiles (Chrome, Chromium, Brave, Edge, Firefox) and audits every installed extension from itsmanifest.json: name, version, permissions, host scope, content scripts, update channel, manifest version. It flags uBlock Origin below 1.76.4 as HIGH. -
filtersparses an ABP-syntax filter file (uBO "My filters" export, AdGuard custom filters). Anytrusted-*scriptlet not in your baseline is HIGH. Broad-scope$redirect/$csprules are HIGH; block-all$document/$popuprules are MEDIUM. Exception rules, cosmetic hiding rules, and ordinary network blocks are never flagged. -
checkis the CI/local gate: scan + filters together, failing (exit 1) only on HIGH findings not covered by the baseline. -
explainshows every evidence item behind one finding, with what to do next. -
initwritesfilterguard.config.jsonand prints the My filters export path for your browser. -
diffcompares two baselines: new / fixed / unchanged findings.
Exit codes: 0 = no HIGH finding, 1 = at least one HIGH finding, 2 = usage error or nothing to audit.
Install from source only - the filterguard package is not published to npm yet (the name is unclaimed; npm returned 404 when checked on 17 November 2026):
git clone https://github.com/techdevfun/filterguard.git
cd filterguard
npm install
npm run build
npm link # or run ./dist/cli.js directly
Export your filter list first (the live list lives in extension storage, not a plain file): uBlock Origin Dashboard -> My filters -> Export; AdGuard Settings -> User rules -> Export. Then:
filterguard init
filterguard check --filters ~/Downloads/my-filters.txt
Example output (synthetic fixture - the demo GIF above is rendered from a real transcript, demo/transcript.txt):
target severity reason
----------------------------------------------------------------------------------------------------------
uBlock Origin 1.75.2 (cjpalhdlnbpafiamejdnhcphjbkeiagm) HIGH uBlock Origin 1.75.2 is VULNERABLE to GHSA-cmqw-9r2w-6wv2
evidence: uBlock Origin 1.75.2; vulnerable <= 1.75.2, fixed 1.75.4, recommended 1.76.4+ (GHSA-cmqw-9r2w-6wv2)
fix: Versions 1.75.2 and lower are vulnerable. Update to 1.76.4 or newer, then run `filterguard filters` - the update does NOT remove a filter the attack already planted.
line 5: example.com#%#//scriptlet('trusted-rpnt', 't... HIGH trusted scriptlet trusted-rpnt runs JavaScript on example.com; planted-filter vector: GHSA-cmqw-9r2w-6wv2
evidence: line 5, sha256 29346360a4e3
fix: Open uBlock Origin Dashboard -> My filters, find line 5, and delete this rule if you did not write it.
summary: HIGH 5 | MEDIUM 2 | LOW 1 | profiles 1 | extensions 2 | rules parsed 5
Why a baseline, not a blocklist
A HIGH score is about capability, not proof of malice. A content blocker legitimately needs broad host access, and advanced users do write trusted-* scriptlets on purpose - uBO documents them for trusted lists. That is what the baseline is for: scan / filters --save-baseline records the state you reviewed, keyed by sha256 of the rule text (whitespace-insensitive) or of the extension + finding. check then fails only on NEW high findings. Never allowlist a rule you do not recognise, and never let the tool delete anything - it reports, you remove.
The extension half has the same shape of problem. The standard free extension auditor, CRXcavator, was killed in 2023 with no replacement. The December 2025 ShadyPanda campaign put spyware into about 4.3M Chrome and Edge installs over seven years. A February 2026 Barracuda report found 99% of employees have browser extensions installed and 52% have more than ten. The manual audit path is chrome://extensions plus reading each manifest by hand.
How it compares
| Tool | What it does | Where filterguard differs |
|---|---|---|
| CRXcavator (Duo Labs / Cisco) | Pulled a CRX apart by ID: permissions, risk tiers, 0-100 score | Dead - killed 2023, no replacement. filterguard is the local successor for the installed-profile half, plus the filter-list half CRXcavator never had. |
| crx-analyzer (PyPI) | Local Python: downloads the CRX by ID, parses manifest permissions, 0-100 score | Store-download only, one extension at a time, no audit of what is actually installed in your profiles, no filter-list parsing. |
| Spin.AI | Enterprise browser-extension security SaaS | $5,000/year minimum, cloud upload of extension inventory. filterguard is free, local, no data leaves the machine. |
| Chrome Extension Security Analyzer (Apify) | Cloud pay-per-use ($0.20/extension): permission tiers, content-script analysis | Cloud service, per-extension, store-fetched. No local profile audit, no filter lists, no baseline/CI gate. |
| Extension Auditor Pro (Chrome Web Store) | An extension that audits installed extensions on-device | Closest rival for the permission half - but it IS an extension (you install more extension code to audit extensions), and it does not parse uBO/AdGuard filter lists. |
| PlayCISO Extension Scanner | Web tool + npm CI gate: live manifest pull from the store by ID | Store-fetched by ID, not an audit of installed profiles; no filter-list parsing; source private, part of a paid platform. |
| uBlock Origin itself (1.76.0+) | Validates filters before the picker creates them | Validates at creation time only. Never audits the My filters list that already exists. filterguard is the audit uBO does not do. |
| toolwatch (same factory, #16) | Snapshots extension/plugin dirs + sha256, diffs drift | Drift detection on file hashes. Does not understand permission semantics, host scope, or filter-list content - a planted filter inside an unchanged file is invisible to it. |
| extverify (same factory, #15) | Pre-install verification: publisher identity, marketplace trust, hash pinning | Pre-install only. Says nothing about what is already installed or already in My filters. |
Where rivals beat us, honestly: crx-analyzer downloads and inspects the actual CRX code (URL extraction from code) - filterguard v0.1.0 reads manifests + filter text only and does not decompile extension code. Spin.AI / PlayCISO cover fleet policy and store-freshness signals filterguard does not. Extension Auditor Pro runs inside the browser and sees live extension state. filterguard v0.1.0 is a static local audit of manifests + filter lists, not a runtime behavioural monitor, and says so on every output.
Honest limits
- Manifest + filter-list audit only. It does not inspect extension code, runtime behaviour, or store-freshness. "No HIGH findings" is not "browser is safe".
- v0.1.0 reads an exported filter file. The live My filters list lives in extension storage (LevelDB/IndexedDB), not a plain file; direct storage read is on the roadmap.
initprints the exact export path. - Firefox profiles are read from
extensions.json, which lists add-ons but not their permissions - those extensions are reported with a LOW "manifest not readable" note and no permission findings. The uBO version gate still applies. - The parser is conservative: unknown rule kinds become counted parse warnings, never silent drops.
- Version guidance is pinned to the check date, 17 November 2026, with sources (advisory + releases page). If a newer patched uBO ships, the recommended version in the code goes stale - re-check the releases page.
- The demo uses a synthetic fixture profile and a synthetic My filters file (the build machine has no real browser profile; none was faked). The planted rule shape is from the public advisory write-ups, on a fixture domain, never a real victim.
Roadmap
- Read live My filters directly from uBO/AdGuard extension storage (LevelDB/IndexedDB) instead of requiring an export.
- CRX code inspection (URL extraction) for the extensions that score HIGH.
- Fleet mode: one baseline per machine, one report for a team.
- Store-freshness signal: flag extensions whose store listing vanished or whose publisher changed.
Source: github.com/techdevfun/filterguard (MIT). All tools in the series: techdev.fun.

Top comments (0)