Skip to content
GeoInt

GeoIntSkill

by marouane53
Released
MIT
Repository
markdown Data & AnalyticsSecurity

Summary

Find where any photo, screenshot, video frame or Street View image was taken: a GeoGuessr-style first read, then OSINT proof with OCR, maps, street-level matching and geometry.

Features

  • Writes a gut-call location first, then proves or overturns it with evidence
  • One-command first pass: metadata, OCR, text-to-country signals, object crops, camera calibration, world prior
  • Candidate board with likelihood ratios capped by evidence type
  • Street-level search across 11 providers and whole-town panorama sweeps with MegaLoc
  • OpenStreetMap, city open data, satellite, terrain and sun geometry
  • Per-photo timestamped archive with report, evidence sheets and scoring

Install This Skill

Add this skill to your favorite AI agent in a few steps.

Any AI agent

This skill is plain instructions — it works with any assistant that accepts custom instructions or system prompts.

  1. Copy the skill content with the button below.
  2. Paste it into your agent's instruction file or system prompt (for example AGENTS.md, .cursorrules, or a custom instructions field).
  3. Ask the agent to apply the skill whenever the task matches.

Skill Content

Markdown Content

Copy this content and use it with your preferred AI agent

# GeoInt

An agent skill that finds where a photo, screenshot, video frame or Street View image was taken, and shows its evidence.

## How it works
1. **Gut call** – the agent writes its own expert read of the scene before running anything.
2. **First pass** – `recon.py` runs metadata, OCR, text-to-country lookups, object crops, camera calibration and a learned world prior in one command.
3. **Candidate board** – `board.py` ranks countries, regions and towns with likelihood ratios capped by evidence type and explicit exclusion rules.
4. **Street level** – `pano.py`, `sweep.py` and `refsheet.py` pull and rank panoramas across 11 providers and match them with MegaLoc.
5. **Geometry** – OpenStreetMap queries, city open data, satellite tiles, terrain and sun angles pin the camera position.
6. **Archive** – each photo gets a timestamped folder with `report.md`, evidence sheets and a score once the truth is known.

## Install
```bash
git clone https://github.com/marouane53/GeoInt.git ~/.agents/skills/geoint
ln -s ~/.agents/skills/geoint ~/.claude/skills/geoint
uv run ~/.agents/skills/geoint/scripts/doctor.py --network
```

Then give your agent a photo and ask: where was this taken?

Usage Instructions

Learn how to use this skill with different AI agents.

Generic Instructions

Clone into ~/.agents/skills/geoint and symlink it into ~/.claude/skills/geoint for Claude Code (Codex and other SKILL.md agents read ~/.agents/skills directly). Requires Python 3.10+, uv and curl; run scripts/doctor.py --network to check the setup.

Description

GeoInt is an agent skill for photo geolocation that pairs the model's own read of a scene with an investigation that has to back it up. The agent writes down a gut call first (soil, vegetation, road paint, poles, plates, scripts, architecture), then proves, narrows or overturns it with web research and 34 local Python tools, down to street or camera level when the evidence allows.

A one-command first pass reads deep metadata, OCR text and text-to-country signals (scripts, phone numbers, domains, postal codes, around 50,000 brands, 2,100 regional plate codes), runs StreetCLIP and GeoCLIP world priors, crops objects with OWLv2 and calibrates the camera with GeoCalib. A reasoning board ranks candidates with likelihood ratios capped by evidence type. Street-level search covers 11 imagery providers and can sweep every panorama in a town with MegaLoc place recognition; OpenStreetMap, city open-data portals, satellite tiles, terrain and sun geometry narrow the rest. Every photo gets a timestamped archive with a report, evidence images and a scoreboard once the true location is known.

GeoInt builds on geo-sleuth (MIT) and credits it line by line: it adds 13 tools, a rewritten method and a worldwide clue library. It needs Python 3.10+, uv and curl, optionally exiftool, ffmpeg and about 6.2 GB of local models; StreetCLIP weights are non-commercial.

Install

Related Skills

See all GeoInt alternatives

Skill: geo-sleuth

by Oldcircle

Featured

Find where a photo was taken: EXIF, OCR, OpenStreetMap, skyline and satellite matching, with coordinates, error radius and evidence images.

Development
1152 views 415 copies

Skill: Hugging Face Vision Trainer

by Hugging Face

Fine-tune object detection, image classification and SAM/SAM2 segmentation models on Hugging Face Jobs cloud GPUs, with dataset validation before you spend GPU minutes.

Development

Skill: NVIDIA TAO RT-DETR

by NVIDIA

Trains, distils, quantises, evaluates, exports and runs inference for RT-DETR real-time object detection in NVIDIA TAO, routing training through AutoML hyperparameter search by default.

Development

Skill: NVIDIA NV-Reason-CXR

by NVIDIA

Runs NVIDIA's NV-Reason-CXR-3B chest X-ray reasoning model through a documented wrapper, locally on GPU or via the public Hugging Face Space, and returns the model's full reasoning trace as engineering evidence.

Development
Browse all skills →