ClawTV

ReviewAudited by ClawScan on May 10, 2026.

Overview

ClawTV appears coherent and purpose-aligned, but it can autonomously control your Apple TV, send TV screenshots to Anthropic, and store TV/Plex credentials locally.

Install only if you are comfortable with an AI-controlled Apple TV remote that can send screenshots to Anthropic and store Apple TV/Plex credentials in plaintext under ~/.clawtv. Review file permissions, use limited API keys where possible, monitor API costs, and avoid running AI mode on sensitive account or personal-information screens.

Findings (4)

Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.

What this means

A mistaken goal or model decision could open apps, type searches, change TV settings, or play content without asking at every step.

Why it was flagged

AI mode can repeatedly decide and execute Apple TV remote actions after a user gives a goal.

Skill content
The `do` command runs in a loop (up to 20 steps) and makes Claude API calls automatically
Recommendation

Use AI mode for clear, limited tasks, monitor the TV while it runs, and use manual commands for sensitive settings or accounts.

What this means

Any local process or user able to read that config file could potentially reuse those credentials to control your Apple TV or Plex setup.

Why it was flagged

The skill stores local credentials that allow Apple TV remote control and, if configured, Plex server/client access.

Skill content
Apple TV pairing credentials and Plex tokens are stored in plaintext in `~/.clawtv/config.json`
Recommendation

Restrict permissions on ~/.clawtv/config.json, only configure Plex if needed, and revoke or rotate tokens if the file is exposed.

What this means

Screenshots may reveal what you watch, search queries, visible account details, or other private TV-screen content to Anthropic's API.

Why it was flagged

The skill sends TV screenshots, the user goal, and conversation history to an external AI provider during autonomous navigation.

Skill content
When using the `do` command ... screenshots are sent to Anthropic's Claude API for vision analysis
Recommendation

Avoid AI mode on sensitive screens, understand Anthropic's data policy, delete local screenshots when appropriate, and monitor API usage/cost.

What this means

The skill may require local automation permissions and can interact with QuickTime/Xcode-related capture workflows on your Mac.

Why it was flagged

The screenshot feature relies on local macOS application automation, which is expected for this Apple TV control use case but should be noticed.

Skill content
Automates the "Take Screenshot" button in Xcode's Devices window via AppleScript
Recommendation

Grant macOS automation/screen-capture permissions only if you trust the skill and understand the screenshot method being used.