pp-scrape-do
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructs the agent to download and install the
scrape-do-pp-clibinary usingnpx -y @mvanhorn/printing-press-libraryorgo install github.com/mvanhorn/printing-press-library/.... These resources originate from the vendor's own package registry and repository. - [COMMAND_EXECUTION]: The skill utilizes the
Read Bashtool to execute a local CLI. It includes a--deliver file:<path>feature that allows writing command output to arbitrary file system paths, which could be misused to overwrite sensitive configuration files or create persistent scripts if not carefully restricted. - [DATA_EXFILTRATION]: The CLI provides a
--deliver webhook:<url>flag, enabling the transmission of scraped data or command results to external HTTP endpoints. Additionally, thefeedback --sendcommand is designed to transmit locally stored feedback to a remote endpoint if configured. - [INDIRECT_PROMPT_INJECTION]: The skill provides a surface for injection through the
sqlcommand, which executes raw SQL strings against a local SQLite database, and thegoogle searchcommand, which processes natural language queries. The instructions lack explicit mention of input sanitization or boundary markers to prevent the agent from being manipulated by malicious search results or untrusted data sources.
Audit Metadata