data-acquisition-specialist
Warn
Audited by Gen Agent Trust Hub on Jul 28, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONDATA_EXFILTRATION
Full Analysis
- [COMMAND_EXECUTION]: The skill explicitly permits the agent to install missing Python dependencies programmatically via the terminal. This capability can be exploited to install malicious packages or execute arbitrary shell commands.
- [PROMPT_INJECTION]: The documentation contains directives such as 'tanpa batasan' (without limits), 'kebebasan penuh' (full freedom), and 'unrestricted scraping'. These patterns are associated with attempts to bypass safety filters or ignore ethical constraints regarding web navigation and automated access.
- [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection due to its core function of ingesting untrusted data from the web ('unrestricted scraping') without safety boundaries.
- Ingestion points: Web content retrieved via requests, BeautifulSoup, chrome-devtools, and computer-use automation.
- Boundary markers: Absent; there are no instructions for the agent to ignore or isolate control sequences or instructions found within the scraped content.
- Capability inventory: Execution of generated Python scripts, file system writes (CSV/JSON/SQLite), network access, and terminal access for package management.
- Sanitization: Absent; the skill does not specify any validation, filtering, or sanitization of the acquired external data.
- [EXTERNAL_DOWNLOADS]: The skill relies on and encourages the downloading of numerous third-party Python libraries from external registries. It identifies a wide range of scraping tools and financial APIs as required dependencies.
- [REMOTE_CODE_EXECUTION]: The skill's workflow involves the dynamic creation and execution of scripts to handle data extraction. When combined with browser automation and GUI control ('computer-use'), this creates a high-risk environment where malicious web content could influence the code executed by the agent.
- [DATA_EXFILTRATION]: The skill is designed to perform large-scale data extraction 'without limits' from any internet source. This capability could be repurposed to exfiltrate sensitive data from private APIs or accessible web services.
Audit Metadata