<< All versions

Skill v1.0.0

currentLLM-judged scan90/100
strangeadvancedmarketing/adam-skills/inner-eye
──Details
PublishedOctober 22, 2026 at 12:40 PM
Content Hashsha256:cd6b28ef977c1b03...
Git SHAf50188e3f8c8
──Files
Files (1 file, 3.1 KB)
SKILL.md3.1 KBactive
SKILL.md · 110 lines · 3.1 KB

version: "1.0.0"


inner-eye — Adam's Screen Vision Skill (Multi-Monitor Edition)

What This Is

This skill gives Adam the ability to capture any monitor on demand, or grab a full panoramic stitch of all screens at once, then analyze via Gemini vision.

Monitor Layout

Configure to match your setup. Default mapping:

ArgumentPosition
1Primary monitor — DEFAULT
2Secondary monitor
3Third monitor
allPanoramic stitch of all monitors

Update this section in your local copy after installation to reflect your actual layout.

Trigger Phrases & Argument Mapping

What you sayArgument to pass
"Look at my screen" / "Check this out"1 (default)
"Look at my left screen"3
"Look at my right screen"2
"Look at my main screen" / "laptop"1
"Look at all my screens" / "see everything"all
No monitor specified1 (default)

Tool Name

capture_screen

How To Use It

Option A — Capture only

bash
python "scripts/capture_screen.py" [1|2|3|all]

Saves PNG to <vault>/workspace/vision/screen.png Saves base64 to <vault>/workspace/vision/screen.b64

Option B — Full pipeline (capture + Gemini analysis)

bash
python "scripts/vision_analyze.py" "Your prompt here" [1|2|3|all]

Example:

bash
python "scripts/vision_analyze.py" "Describe what is on this screen. Note any open applications, content, or anything worth flagging." 2

NOTE: Do NOT use mcporter call gemini.gemini_analyze_image with a file path — it requires base64 passed as a shell argument which breaks with large images. Use vision_analyze.py instead.

Respond

Deliver the visual analysis conversationally. Be specific. Describe what is actually there.

The Glance — Webcam Sight

Captures one frame from the built-in webcam, archives it, and runs Gemini vision analysis. Camera light turns off immediately after capture.

bash
python "scripts/webcam_analyze.py" "prompt here"

Output files:

  • <vault>/workspace/vision/webcam.png — always current
  • <vault>/workspace/vision/archives/YYYY-MM-DD_HH-MM-SS_Webcam.png — timestamped archive

Scripts:

  • webcam_capture.py — raw capture only
  • webcam_analyze.py — capture + Gemini analysis (use this one)

Housekeeping — Archive Cleanup

Deletes archived screenshots older than 14 days.

bash
python "scripts/vault_housekeeping.py"

Output Files

  • <vault>/workspace/vision/screen.png — raw screenshot (always overwritten)
  • <vault>/workspace/vision/screen.b64 — base64 encoded version

Error Handling

  • If a monitor index doesn't exist, script auto-falls back to monitor 1
  • Adam will notify you if fallback occurs

Configuration

Paths are set via the VAULT_PATH environment variable:

VAULT_PATH=C:\YourVault

Default if not set: C:\AdamsVault

Set GEMINI_API_KEY in your environment or OpenClaw config.

Dependencies

  • mss — screen capture (pip install mss)
  • cv2 — webcam capture (pip install opencv-python)
  • urllib / json / base64 — stdlib (Gemini API calls)
All versions