duckduckgo-search
Warn
Audited by Gen Agent Trust Hub on Sep 28, 2026
Risk Level: MEDIUMREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONMETADATA_POISONING
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill instructs the agent to install a package named 'ddgs' via pip. Although a package by this name exists on PyPI, it is not the standard 'duckduckgo-search' library that provides the 'ddgs' CLI tool. Recommending an unverified package under the claim of a package name change introduces a significant supply chain risk where malicious code could be installed.
- [COMMAND_EXECUTION]: The skill provides instructions and a shell script for executing various CLI commands via the terminal to install software and perform web searches.
- [INDIRECT_PROMPT_INJECTION]: The skill fetches and processes external search results from DuckDuckGo, which are controlled by third parties. 1. Ingestion points: Data enters the context through CLI output and Python API results in SKILL.md. 2. Boundary markers: None present to delineate or warn against untrusted content. 3. Capability inventory: Uses terminal and execute_code tools. 4. Sanitization: No sanitization or filtering of the search snippets is performed before they are processed by the agent.
- [METADATA_POISONING]: The skill metadata and documentation contain a misleading claim that the 'ddgs' package is the new name for 'duckduckgo-search', which could deceive agents or users into installing incorrect or unverified software.
Audit Metadata