Security audit
Security checks for vulnerabilities and agentic risk
This skill is a scraping guide with useful compliance checks, but it also teaches anti-bot evasion, Cloudflare bypass approaches, stealth browser configuration, and CAPTCHA-solving services.
Install only if you need a scraping methodology and can enforce your own authorization boundaries. Use it for sites you own, have permission to test, official APIs, or clearly permitted public data; avoid using its stealth, proxy, Cloudflare, or CAPTCHA guidance to bypass site protections or terms.
The skill presents anti-detection as a normal step in a recommended scraping workflow: assess legality, choose tooling, then defeat anti-bot detection. That sequencing operationalizes evasion behavior and makes defensive bypass appear endorsed as part of standard usage, which can facilitate unauthorized scraping at scale.
The README explicitly promotes 'defeats anti-bot detection' using proxy rotation, fingerprint diversity, and stealth patterns, but does not pair that guidance with clear restrictions about authorization, terms-of-service, or account and legal risk. In a scraping skill, this materially increases the chance the skill will be used to evade website protections rather than for compliant data collection.
The phrase 'defeats anti-bot detection' uses product-style language to normalize evasion of website protections. Even without exploit code, this kind of framing encourages users to treat bot-detection bypass as a legitimate capability, which is dangerous in a skill explicitly designed to help automate scraping.
Listing 'anti-detection strategies' and 'Cloudflare bypass' among core included features advertises website-protection circumvention as a supported capability. In this context, the skill becomes more dangerous because it is a packaged methodology for production-scale scraping, making misuse easier and more scalable.
The manifest description is extremely broad and can trigger on many ordinary requests involving scraping, competitor monitoring, or automation. Overbroad activation increases the chance the skill is invoked in contexts where legal, safety, or authorization constraints have not been established, making risky anti-detection and bypass guidance more readily available.
Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.
| Public data, no login, robots.txt allows | LOW | hiQ v. LinkedIn (2022) |
| Public data, robots.txt disallows | MEDIUM | Meta v. Bright Data (2024) |
| Behind authentication | HIGH | Van Buren v. US (2021), CFAA |
| Personal data without consent | HIGH | GDPR Art. 6, CCPA §1798.100 |
| Republishing copyrighted content | HIGH | Copyright Act §106 |
| Price/product comparison | LOW | eBay v. Bidder's Edge (fair use) |
| Academic/research use | LOW-MEDIUM | Varies by jurisdiction |
The request header example fixes 'Accept-Language' to 'en-US,en;q=0.9', which imposes a specific language/locale preference in natural-language configuration. Because the document does not present this as optional or user-selected in that example, it may violate language/locale policy expectations.
The Playwright stealth configuration forces locale, timezone, geolocation, and permissions to impersonate a specific regional user profile, and it is presented under 'Anti-Detection & Stealth.' In context, this is not merely localization guidance; it is operational advice for masking automation and evading anti-bot controls, which increases the skill's misuse potential.
The skill explicitly recommends CAPTCHA-solving services such as 2Captcha and Anti-Captcha as a fallback, which materially escalates the skill from ordinary scraping guidance into anti-abuse control bypass. Even though the surrounding text says to slow down first, including named challenge-bypass services provides actionable assistance for defeating site protections and can facilitate unauthorized collection at scale.
The natural-language commands are open-ended and directly support tasks like building scrapers, diagnosing blocks, extracting data from arbitrary URLs, and finding hidden APIs. Without scope gates for authorization and lawful use, these commands lower friction for misuse and may accidentally activate in high-risk contexts.
No suspicious patterns detected.