Back to skill
Skillv1.0.0
ClawScan security
Awesome GeeLark Skill · ClawHub's context-aware review of the artifact, metadata, and declared behavior.
Scanner verdict
SuspiciousApr 29, 2026, 7:39 AM
- Verdict
- suspicious
- Confidence
- medium
- Model
- gpt-5-mini
- Summary
- The skill generally matches its stated purpose (GeeLark cloud-phone + RPA automation) but contains several inconsistencies and operational instructions that increase risk (missing declared credentials/binaries, runs local ADB subprocesses, asks to store tokens locally).
- Guidance
- What to consider before installing: - Functionality: The code and docs match the description: this is a GeeLark cloud-phone manager that can boot devices, enable ADB, run uiautomator2 automation, and submit RPA tasks that may include social-account credentials. That behavior is consistent with the stated purpose. - Inconsistencies to fix/verify: The registry metadata declares no required environment variables or required binaries, but the skill expects an API token/appId/apiKey stored in assets/config.json and uses the adb binary/uiautomator2. Confirm that you will provide and protect those credentials and ensure adb is available/expected on the host. - Sensitive data: init_config.py creates assets/config.json containing API keys/tokens. Review init_config.py to confirm where/how credentials are stored and that file permissions/ignore rules are acceptable. RPA tasks may require social platform usernames/passwords — avoid feeding long-lived account credentials to any agent unless you fully trust the code and runtime environment. - Local command execution: The skill runs adb via subprocess (including a required 'glogin' step) and instructs pip installing uiautomator2 with --break-system-packages. Run this in an isolated/sandbox environment (VM/container) first and verify behavior; do not run on sensitive hosts. - Trust and provenance: Source/homepage is unknown in the registry entry. Although the docs reference openapi.geelark.com and a GitHub URL, verify the repository and the authorship before use (confirm the repository matches package files and the API base URL). If you cannot verify the origin, be cautious. - Recommended actions before use: (1) Inspect scripts/init_config.py, geelark_client.py, and any logging code to see exactly what data is stored/transmitted; (2) run the skill in a sandboxed VM; (3) avoid giving the agent unattended autonomous invocation until you confirm safe behavior; (4) only provide short-lived API keys or limited-scope credentials when possible; (5) ensure logs do not leak full secrets and review log-masking behavior. If you want, I can: point out where credentials are loaded in the code, summarize the exact subprocess calls that run adb/glogin, or suggest concrete sandboxing steps to test safely.
Review Dimensions
- Purpose & Capability
- noteName/description, docs, and code files align: the package is a GeeLark Cloud Phone client + helpers for booting devices, ADB/uiautomator2 automation, and RPA tasks for social platforms. However the registry metadata declares no required credentials or binaries even though the code/docs expect an API token/appId/apiKey (assets/config.json) and use local adb/uiautomator2. The missing declarations are an incoherence worth noting.
- Instruction Scope
- concernSKILL.md instructs the agent/operator to run scripts/init_config.py (creates assets/config.json containing API credentials), to pip-install uiautomator2, and to execute ADB commands via subprocess (including a 'glogin' step) and use uiautomator2 to dump UI hierarchies and perform taps/inputs. It also shows RPA tasks that include third-party account credentials (social platform usernames/passwords) and file paths. These instructions access and store sensitive credentials and run local system commands — beyond simple REST API calls — increasing attack surface and risk of accidental credential exposure or abuse. The SKILL.md also auto-loads credentials from a file not declared in the registry metadata.
- Install Mechanism
- noteThere is no formal install spec in the registry. The documentation tells users to pip install uiautomator2 (with --break-system-packages). That is a normal Python dependency for device automation but may require native system changes; using --break-system-packages can be intrusive. No downloads from external personal servers are present in the manifest. Overall install risk is moderate and centered on the uiautomator2 requirement and local ADB usage.
- Credentials
- concernThe skill expects API credentials (token, appId, apiKey) stored in assets/config.json and will log operations locally; it also accepts RPA payloads containing third-party account credentials and proxy information. Yet the registry entry lists no required env vars or required binaries (e.g., adb) — an inconsistency. Requesting or storing multiple credentials (GeeLark API keys plus social account passwords and proxy credentials) is proportionate to the functionality but must be declared clearly and handled cautiously; the current packaging omits that declaration and thus looks under-specified and risky.
- Persistence & Privilege
- noteThe skill is not marked always:true and does not request elevated system privileges. It writes assets/config.json and logs to logs/cloudphone/ (normal for a client). However, default platform behavior allows autonomous invocation; combined with this skill's access to API tokens and potential to run local adb/subprocess commands, autonomous operation increases blast radius — test in a controlled environment and consider restricting autonomous runs until validated.