🦌 Wildlife Census — OpenClaw Skill

Track every critter that visits your yard. A lightweight species logging and field journal system for any OpenClaw agent on macOS.

Manual logging, camera-based auto-detection via VLM, species statistics, first sighting tracking, weekly trend reports, and searchable history — all stored locally as JSONL files.

---

What It Does

• Log sightings manually or automatically via camera + VLM
• Track species with counts, behaviors, timestamps, and first-sighting flags
• Weekly reports with trends — "Hummingbird visits peaked at 47, up from 31 last week"
• Life list — every species you've ever recorded, with first sighting dates
• Search history — "Have we seen any hawks this month?"
• Field journal — generate narration-ready summaries for daily briefings

Everything is stored as local JSONL files. No cloud. No database. No dependencies beyond bash and basic Unix tools.

Installation

git clone https://github.com/MS-707/wildlife-census-skill.git
cd wildlife-census-skill
chmod +x census.sh

For OpenClaw integration, copy or symlink to your skills directory:

cp -r . ~/.openclaw/skills/wildlife-census/

Usage

./census.sh <command> [args]

Command	What It Does	Example
log <species> [count] [notes]	Record a sighting	./census.sh log "Scrub Jay" 2 "at the feeder"
snap	Camera capture → VLM identify → auto-log	./census.sh snap
today	Today's sightings summary	./census.sh today
week	7-day summary with trends	./census.sh week
stats	All-time species statistics	./census.sh stats
search <species>	Search sighting history	./census.sh search hawk
firsts	All first-time species sightings	./census.sh firsts
journal	Generate field journal data	./census.sh journal

Data Storage

All data lives in ~/.openclaw/workspace/wildlife/:

wildlife/
├── sightings.jsonl    ← Every sighting, one JSON object per line
├── firsts.jsonl       ← First-time species records
└── species.json       ← Species database with cumulative stats

Each sighting record:

{
  "timestamp": "2026-03-22T15:24:01-07:00",
  "species": "California Scrub-Jay",
  "count": 1,
  "behavior": "at the feeder",
  "source": "manual",
  "first_sighting": false
}

Integration with Other OpenClaw Skills

Bird Watcher

The bird-watcher-skill automatically logs detections to the wildlife census when installed. Every YOLO + Moondream detection becomes a sighting record.

Camera / VLM

The snap command captures a photo and sends it to a local VLM (Moondream Station on localhost:2020) for species identification, then auto-logs the result. Requires a webcam and Moondream running.

Morning Briefings

Your OpenClaw agent can pull today or week summaries into daily briefings — "Three new species this week, hummingbird visits up 40%."

What This Skill Cannot Do

• It's not an automated monitoring system. For real-time detection, use the bird-watcher-skill. The census is a logging and reporting tool.
• The snap command requires Moondream Station running locally. Without it, use manual log commands.
• Species identification accuracy depends on the VLM. Common backyard wildlife at close range works well. Distant or unusual species may get generic labels.

Requirements

• macOS (uses bash, date, jq-style JSON handling)
• Moondream Station (optional, for snap auto-detection)
• No Python dependencies — pure bash + standard Unix tools

Privacy

• All data stored locally as flat files
• No cloud services, no accounts, no telemetry
• Camera access (for snap) is gated by macOS permissions
• Your wildlife data never leaves your machine

Credits

• Created by Mark Starr & Victor as an OpenClaw skill
• Designed for Sonoma County, CA but works anywhere
• Pairs with bird-watcher-skill for automated detection

License

MIT — see LICENSE for details.