screen-describe
Skill0 starsChanged 2 months agoApache-2.0View on GitHub
Ask your local Claude what's on your screen. Captures the macOS screen, hands the image to your own `claude` CLI (your Pro/Max, no API key — BYOK), prints a short description, and deletes the screenshot immediately. Read-only — it looks, never controls. Use when the user says "what's on my screen", "describe my screen", "what am I looking at", or wants an AI second pair of eyes on their display.
In this repository17 other files it gives its agents.
- aria-skill-candidatesSkill
- aria-skill-templateSkill
- aria-skill-testSkill
- call-scrubSkill
- cloudflare-dns-deploySkill
- doc-to-pdfSkill
- ffmpeg-audioSkill
- github-repo-deploySkill
- goal-to-planSkill
- headless-claudeSkill
- live-code-watchSkill
- premortemSkill
- session-cabinetSkill
- skill-manifest-genSkill
- sprint-scaffoldSkill
- surgical-patchSkill
- video-ingestSkill
The file
Files
screen-describe
Your local Claude, as eyes on your screen. BYOK, read-only, screenshot deleted after.
python3 screen_describe.py # dry pre-flight (checks deps, captures nothing)
python3 screen_describe.py --apply # capture + describe what's on screen
python3 screen_describe.py --apply --question "is there an error visible anywhere?"Why
Sometimes you want an AI to actually see your screen — "what's this error", "what
am I looking at" — without screenshotting + uploading anywhere. This runs entirely on
your machine: screencapture → your local claude (your own plan) → a 2-3 sentence
read. The screenshot is deleted the moment it's read; nothing leaves your box except
the model call you already pay for.
Behavior
- Default: dry pre-flight — confirms
screencapture+claudeare present, exits 0. --apply: capture, describe, delete the shot.- Missing deps / no Screen Recording permission → clear
FATAL+ exit 2.
Requires
- macOS (
screencapture) + Screen Recording permission for the running app. - The
claudeCLI, logged in to your Anthropic plan. No API key, no Python packages.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.