Skip to content

Add Claude Code skill for desktop GUI automation#1693

Open
tdimino wants to merge 1 commit intoopeninterpreter:mainfrom
tdimino:community/claude-code-skill
Open

Add Claude Code skill for desktop GUI automation#1693
tdimino wants to merge 1 commit intoopeninterpreter:mainfrom
tdimino:community/claude-code-skill

Conversation

@tdimino
Copy link

@tdimino tdimino commented Feb 23, 2026

Summary

Adds a community-contributed Claude Code skill in .claude/skills/open-interpreter/ that wraps Open Interpreter's Computer API for desktop GUI automation. No changes to Open Interpreter's source code or package.

What is this?

Claude Code reads .claude/skills/ directories for specialized capabilities (similar to how .github/ configures GitHub Actions). This skill provides standalone Python scripts that wrap OI's pyautogui + pytesseract primitives — screenshot capture, mouse clicking (by coordinates or OCR text), keyboard input, and screen text detection.

Three integration modes:

  • Library: Claude Code reasons from screenshots, dispatches actions via the bundled scripts
  • OS subprocess: delegates entire GUI tasks to OI's --os agent loop
  • Local agent: offline computer use via Ollama

Why contribute this upstream?

Open Interpreter gave us the foundation — the Computer API, OS Mode, and the pyautogui/pytesseract integration are all OI's work. This skill just makes that functionality accessible to Claude Code users working with the OI codebase. Felt right to offer it back rather than keep it in a separate repo.

What's included

.claude/skills/open-interpreter/
├── SKILL.md                     # Skill instructions (Claude Code reads this)
├── README.md                    # Human-readable docs
├── scripts/                     # 8 standalone Python/Bash scripts
│   ├── oi_screenshot.py         # Screen capture with Retina metadata
│   ├── oi_click.py              # Click by coordinates or OCR text
│   ├── oi_type.py               # Keyboard input, hotkeys
│   ├── oi_find_text.py          # OCR screen reading → JSON
│   ├── oi_computer.py           # Unified dispatch
│   ├── oi_os_mode.py            # Managed OI subprocess
│   ├── oi_permission_check.py   # macOS permissions check
│   └── oi_install.sh            # One-shot install
└── references/                  # 3 reference docs

Test plan

  • All scripts tested on macOS (Retina display, Accessibility + Screen Recording permissions)
  • No changes to interpreter/ package or tests/
  • Quality reviewed — 7 fixes applied (AppleScript injection sanitization, zombie process cleanup, Retina coordinate precision, flag conflict resolution)

Built with Claudicle, an open-source soul agent framework.

Add a community-contributed Claude Code skill that wraps Open Interpreter's
Computer API (pyautogui, pytesseract) for desktop GUI automation. Provides
standalone scripts for screenshot capture, mouse clicking (by coordinates or
OCR text), keyboard input, and screen text detection.

Three integration modes:
- Library: Claude Code reasons from screenshots, dispatches actions via scripts
- OS subprocess: delegates entire GUI tasks to OI's --os agent loop
- Local agent: offline computer use via Ollama

No changes to Open Interpreter's source code or package.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant