parallel-web-search
Pass
Audited by Gen Agent Trust Hub on Jun 27, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to construct and execute a shell command (
parallel-cli search "$ARGUMENTS") incorporating user-supplied input directly. This pattern creates a command injection surface if the user-provided string contains shell metacharacters that are not properly sanitized or escaped by the execution environment. - [PROMPT_INJECTION]: The skill processes untrusted data from the web, creating an indirect prompt injection surface.
- Ingestion points: External content is retrieved via
parallel-cliand saved to/tmp/$FILENAME.jsonfor processing by the agent. - Boundary markers: Absent. The instructions do not use delimiters to wrap external content or provide explicit directives to the agent to ignore instructions embedded within the retrieved search results.
- Capability inventory: The skill has capabilities for shell command execution (
parallel-cli) and local file system read/write access (/tmp/). - Sanitization: Absent. While the skill instructs the agent to filter out "navigation noise," there is no evidence of sanitization or validation of the actual content to prevent malicious instructions from influencing the agent's behavior.
- [EXTERNAL_DOWNLOADS]: The skill performs network operations to fetch data from the internet. This is the intended purpose of the skill and utilizes the vendor's (
parallel-web) own CLI tool and infrastructure.
Audit Metadata