ClawHub

adb-phone-control

v1.0.1

Use when the user asks to control, operate, or automate an Android phone via ADB — tapping, swiping, typing, launching apps, or any UI interaction on a conne...

⭐ 0· 95·0 current·0 all-time
by@txmonkey
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
VirusTotalVirusTotal
Benign
View report →
OpenClawOpenClaw
Benign
high confidence
✓
Purpose & Capability
Name/description (ADB control) matches the code and SKILL.md: both require adb and python3 and implement UI dump, screenshot, tap/swipe/input, and an app explorer. No unrelated binaries or credentials are requested.
ℹ
Instruction Scope
SKILL.md and helper scripts instruct the agent to dump UI, pull screenshots/dumps to local path, and run automated taps/swipes/inputs. This is expected for device automation, but the app_explorer recursively clicks elements which can trigger side effects (sending messages, purchases, or destructive actions). The instructions also tell the agent to 'read' screenshots for verification — that will expose device screen content to whatever model/tool is used to view images.
✓
Install Mechanism
Instruction-only install (no download/extract). Two local code files are included and executed by sourcing/running them. No external installers or remote downloads are used by the skill itself.
✓
Credentials
No secret environment variables or credentials are requested; ADB_OUTPUT_DIR is optional and reasonable. The skill does not ask for cloud keys or unrelated secrets.
✓
Persistence & Privilege
always is false and the skill does not request persistent system-wide privileges. It operates only by invoking adb commands against a connected device and writing output into a local output directory.
Assessment
This skill appears to do what it claims (ADB-based UI automation). However: 1) automated exploration will click/tap elements and can perform destructive or privacy-sensitive actions on the device (send messages, change settings, make purchases). Only run it on a test device or after explicitly reviewing which package and actions will be targeted; prefer manual observe-first runs. 2) Review the included scripts (adb-helpers.sh and app_explorer.py) before use to confirm no unexpected commands for your environment. 3) Keep ADB enabled only when needed and disconnect when done. 4) If you intend to let an agent run this autonomously, restrict it to devices you control and consider disabling autonomous invocation until you’ve tested behavior interactively.

Like a lobster shell, security has layers — review code before you run it.

latestvk975rt5c2fastesndp49610c5d84182p

License

MIT-0
Free to use, modify, and redistribute. No attribution required.
Termshttps://spdx.org/licenses/MIT-0.html

Comments