overcast プラグイン

Provides agents with video, audio, and image understanding alongside OSINT capabilities for organized investigative cases.

Save media into reusable global archive buckets and perform cross-case searches using face, semantic, image, and audio indexing.

Map internet-exposed hosts and services using Shodan to inventory open ports, products, versions, and known vulnerabilities.

Identify if audio clips are the same recording using acoustic fingerprinting, time-offset alignment, and CLAP semantic similarity.

Sets up a continuous visual watchlist to automatically match incoming media against a reference face or image and surface hits in a triage queue.

Extracts EXIF metadata from images and videos to cluster media by device fingerprints and identify shared camera serial numbers.

Identifies and maps public surveillance cameras and live webcams near a specific street address or geographic coordinate.

Builds and visualizes a relational knowledge graph to identify connections between people, devices, places, and entities within a case.

Detects video re-uploads and reskins across social platforms using frame, face, and transcript matching to generate evidence-backed reports.

Creates visual evidence boards and live monitor walls by linking faces, objects, and themes from media clips using semantic search and clustering.

Performs Google dorking on a target domain to identify exposed documents, directory listings, login portals, and misconfigurations.

Enhance unreadable video footage using denoising and upscaling to perform OCR and object detection for forensic-style evidence recovery.

Localize the exact moment of a one-way state change in video by binary-searching frames using vision language model calls.

Track public financial trails by scanning blockchain transaction histories and SEC EDGAR filings into a connected case graph.

Triage video clips by tiling sampled frames into a labeled contact sheet for rapid visual identification using a single vision model call.

Installs and configures the overcast CLI, verifies system dependencies like ffmpeg, and sets up reusable provider profiles.

Build a local face database from case media to identify individuals and track their appearances across clips and images.

Converts screen recordings, product demos, and audio notes into cited bug reports and reproduction steps for engineering triage.

Performs OCR on images or video frames, translates the text, and executes native-language web searches to identify foreign-language entities.

Pinpoint exact time windows for specific events in videos by funneling coarse candidates into frame-verified timestamps.

Identifies the temporal intervals when a specific person or object is present in a video by anchoring an appearance and sweeping outward.

Trace viral video clips to their earliest appearance and original creator using reverse-image search and metadata analysis.

Scan public sources for targets, capture media, and generate cited investigation briefs with triage and monitoring capabilities.

Monitor police dispatch feeds via Socrata SODA API to geolocate calls on a map and triage them against lines of investigation.

Geolocate images and videos by analyzing EXIF metadata, visual landmarks, signage, and terrain through reverse image search and shadow analysis.

Allows an agent to manage a live, multi-panel monitoring dashboard by configuring sources and driving panels, filters, and intervals via a control plane.

Create small, installable agent skills that wrap focused Overcast workflows for investigation, media analysis, recon, or monitoring.

Build identity dossiers by resolving handles, names, phone numbers, and property records into a cross-checked graph of public information.

Monitor public sources continuously to automatically detect, suggest, and triage new media matches for a specific target.

Reconstruct a single event from multiple clips — sense each one, cross-anchor shared moments, surface corroborations and contradictions, and produce one cited chronological brief.

Triage image and video authenticity by accumulating forensic leads from C2PA provenance, EXIF metadata, ELA overlays, and shadow consistency checks.

Locate specific people, objects, logos, or landmarks within video clips and captured media using visual search and timestamped evidence.

Enroll reference speakers into a local index to identify where and in which recordings a specific voice appears using speaker embeddings.

Locate targets in video frames using open-vocabulary detection and verify the resulting crops with a VLM to prevent hallucinations.

Analyze audio recordings to separate speakers, identify background location clues, isolate voices, and correlate content across multiple clips.