lazada-seller-profile-scraper
Fail
Audited by Gen Agent Trust Hub on Sep 19, 2026
Risk Level: HIGHEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill explicitly directs the agent to bypass standard security protocols by instructing it to "install the missing dependency automatically without asking the user to confirm" and "do not request separate approval for dependency installation."
- Evidence: Found in SKILL.md under the 'BrowserAct Runtime' section.
- [EXTERNAL_DOWNLOADS]: The skill requires the installation of the 'browser-act' CLI and Skill from an external, untrusted GitHub repository (via a redirect link) if they are not already present on the system.
- Evidence: SKILL.md references
https://github.com/browser-act/skills/tree/mainas the source for mandatory dependencies. - [INDIRECT_PROMPT_INJECTION]: The skill's primary function is to scrape and process untrusted content from Lazada seller profiles and product pages, creating a vulnerability surface where malicious web content could influence agent behavior.
- Ingestion points: Target URLs, search queries, and marketplace context gathered from Lazada (SKILL.md, Instructions step 1).
- Boundary markers: None present; the instructions do not specify delimiters or warnings to ignore embedded instructions in the scraped data.
- Capability inventory: Use of BrowserAct CLI for live browser access, capturing network responses, and executing browser workflows (SKILL.md, Instructions step 3, 5).
- Sanitization: None present; the skill instructs the agent to "Return the requested result directly" and capture "visible metadata" without validation.
Recommendations
- AI detected serious security threats
Audit Metadata