jhpacker/slopguard

How jhpacker/slopguard works

SlopGuard is a local-only browser extension that scans eligible page images on user request (or a single right-clicked image), runs provenance and visual-detection tiers, and marks the result in the page. The main runtime path is page scan → background detection cascade → offscreen C2PA and model inference → page overlays. Metadata and byte checks run in the background; SynthID preprocessing supports its model input. The excerpts do not show the full cascade or toolbar/context-menu dispatch wiring, so those connections are limited to documented behavior and visible message calls.

Main entry points: background.js Detection coordinator · content.js Image scan and overlays · offscreen.html Offscreen detector host · offscreen.js SynthID watermark inference

Components 16

Page scanning

Image scan and overlays src/content.js
Overlay styles src/overlay.css

Detection pipeline

Detection coordinator src/background.js
Image bytes src/background.js
EXIF/IPTC/XMP checks src/background.js
Byte marker checks src/background.js

Local inference

Offscreen detector host src/offscreen.html
C2PA manifest checks src/offscreen.js
SynthID watermark inference src/offscreen.js
SynthID preprocessing src/synthid-preprocess.js
Visual classifier inference src/offscreen.js
Local ONNX models

Extension settings

Debug options src/options.js
Extension settings storage

Other components

Extension user
Web page

Connections 20

Extension user → Image scan and overlays: requests scan
Web page → Image scan and overlays: provides images
Image scan and overlays → Detection coordinator: requests detection
Detection coordinator → Image bytes: fetches image
Detection coordinator → EXIF/IPTC/XMP checks: parses metadata
Detection coordinator → Byte marker checks: scans bytes
Detection coordinator → Offscreen detector host: dispatches checks
Offscreen detector host → C2PA manifest checks: runs manifest check
Offscreen detector host → SynthID watermark inference: runs watermark check
Offscreen detector host → Visual classifier inference: runs classifier
SynthID watermark inference → SynthID preprocessing: builds model input
SynthID watermark inference → Local ONNX models: uses SynthID model
Visual classifier inference → Local ONNX models: uses OpenFake model
Detection coordinator → Image scan and overlays: returns verdict
Image scan and overlays → Web page: marks images
Overlay styles → Web page: styles overlays
Debug options → Extension settings storage: reads and writes
Detection coordinator → Extension settings storage: reads debug settings
Image scan and overlays → Extension settings storage: reads debug setting
Offscreen detector host → Extension settings storage: reads debug setting

More JavaScript diagrams: tetratensor/mt4-mt5-trade-copier-backend · thisisankit27/f-tree · wenhaoqiasd/gatsby-netlifycms · xyron24/nexeditor · yuukiofuyu/serverless-temp-file-sharing

Drawn by GitDiagram on September 29, 2026 from the repository's file tree, README and source files. Read as Markdown ·