parallel-web-search

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to construct shell commands by interpolating user-provided query strings and agent-generated filenames directly into the parallel-cli tool. This pattern relies on the agent to correctly sanitize inputs to prevent command injection via shell metacharacters.
  • [INDIRECT_PROMPT_INJECTION]: The skill fetches and processes untrusted data from the web, creating a potential surface for indirect instructions to influence the agent's behavior.
  • Ingestion points: External web content is retrieved via the parallel-cli search command and stored in /tmp/ JSON files for processing in SKILL.md.
  • Boundary markers: The instructions do not define delimiters or specific 'ignore' directives to help the agent distinguish between its instructions and the retrieved web content.
  • Capability inventory: The skill has access to the parallel-cli tool (search, balance management), the ability to execute shell commands, and filesystem write access in /tmp/.
  • Sanitization: There is no explicit validation or sanitization logic applied to the external web results before the agent parses and synthesizes them into a response.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 11:47 AM
Security Audit — agent-trust-hub — parallel-web-search