result

Pass

Audited by Gen Agent Trust Hub on Sep 14, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted external data which could lead to indirect prompt injection.
  • Ingestion points: The output of the parallel-cli research poll command enters the agent's context at runtime.
  • Boundary markers: Absent. There are no delimiters or explicit instructions telling the agent to ignore any potential command instructions embedded within the research results.
  • Capability inventory: The skill uses the Bash(parallel-cli:*) tool to execute commands.
  • Sanitization: Absent. The research results are presented directly to the agent without escaping or structural validation.
  • [COMMAND_EXECUTION]: User-provided input via $ARGUMENTS is directly interpolated into a shell command line (parallel-cli research poll "$ARGUMENTS" --json). While enclosed in double quotes to mitigate word-splitting, it could still allow command or parameter expansion depending on how the underlying execution environment handles shell evaluation.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 14, 2026, 08:07 PM
Security Audit — agent-trust-hub — result