Back to skill
Skillv1.0.0
ClawScan security
Car Scraper · ClawHub's context-aware review of the artifact, metadata, and declared behavior.
Scanner verdict
BenignMar 9, 2026, 8:06 AM
- Verdict
- benign
- Confidence
- high
- Model
- gpt-5-mini
- Summary
- The code, SKILL.md and files are internally consistent with a web-scraping / anti-detection toolset for the named car sites; it will perform network requests and write output files, and includes explicit anti-scraping/evasion techniques — review legal/ethical and runtime impact before use.
- Guidance
- This skill appears to be what it claims (a multi-site car-data scraper with anti-detection features). Before installing/run it: 1) Be aware it will perform outbound HTTP(S) requests to the target sites and may be used to circumvent anti-scraping measures (UA rotation, header fingerprinting, cookie handling, proxy support). Ensure you have the legal right and comply with the target sites' terms of service and local laws. 2) Run in a controlled environment (sandbox or isolated VM) if you're unsure, since the skill will write output files (config.py creates an output/ and openclaw/ directory inside the package). 3) Inspect and sanitize config.py before use — remove or review any proxy entries (they may contain credentials) and set acceptable rate limits to avoid causing harm. 4) Monitor network activity and logs during initial runs. 5) If you need to avoid evasion capabilities, remove/disable anti_detect features (UA rotation, fingerprinting) so behavior is conservative and transparent.
Review Dimensions
- Purpose & Capability
- okName/description (car data scraping from 大搜车/懂车帝/汽车之家) match the provided Python modules and SKILL.md. The scrapers, data model, and OpenClaw export code correspond to the stated capabilities.
- Instruction Scope
- noteSKILL.md instructs the agent to import scrapers and export functions only (no broad file reads). The code will perform network requests to target sites, manage cookies, rotate UAs, use rate limiting and optionally proxies. One side-effect: config.py creates output directories at import time (os.makedirs on package path). The anti-detection logic explicitly aims to evade site protections — functionally coherent but worth noting for policy/ethical review.
- Install Mechanism
- okNo install spec; this is instruction + code files only. Dependencies appear to be standard Python libs (requests, bs4) referenced in requirements.txt — no unknown remote download or extract steps detected.
- Credentials
- okThe skill does not request environment variables, credentials, or external config paths. It contains an (empty) proxy list and configurable headers in config.py; no secrets are required by default. If you add proxies with credentials, that would raise additional risk.
- Persistence & Privilege
- okalways:false and no code that attempts to modify other skills or global agent configuration. The only persistent filesystem action is creating output directories within the package path and writing scraped output (OpenClaw exports), which is expected for this purpose.