parallel-search

Fail

Audited by Gen Agent Trust Hub on Apr 7, 2026

Risk Level: HIGHREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill provides instructions to install its required CLI tool by piping a script from a remote URL (https://parallel.ai/install.sh) directly into a bash shell. This is an insecure installation pattern that executes opaque remote code.
  • [EXTERNAL_DOWNLOADS]: The skill fetches an installation script and potentially other dependencies from the parallel.ai domain during setup.
  • [COMMAND_EXECUTION]: The skill relies on executing shell commands, including curl, bash, and the parallel-cli tool, to perform installation and search tasks.
  • [PROMPT_INJECTION]: The skill ingests untrusted data from the web that is then processed by the agent, creating a surface for indirect prompt injection.
  • Ingestion points: Intelligent excerpts and ranked results from web searches are retrieved via the parallel-cli and passed to the agent context.
  • Boundary markers: The instructions do not define clear delimiters or warnings to prevent the agent from following instructions embedded within the search results.
  • Capability inventory: The skill uses the parallel-cli subprocess and suggests using sessions_spawn to synthesize summaries, which involves tool-calling based on untrusted data.
  • Sanitization: There is no evidence of sanitization or filtering of the content retrieved from the web before it is presented to the LLM.
Recommendations
  • HIGH: Downloads and executes remote code from: https://parallel.ai/install.sh - DO NOT USE without thorough review
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Apr 7, 2026, 11:33 AM
Security Audit — agent-trust-hub — parallel-search