rt-journal-match

Pass

Audited by Gen Agent Trust Hub on Aug 24, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill requires the agent to execute a local script named tools/match_venues.py via python3 using a command template that incorporates user-supplied data.
  • Evidence: python3 tools/match_venues.py --title "..." --abstract "..." --discipline economics/labor --lane empirical --top 15 in SKILL.md.
  • [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface through its method of handling untrusted user data for tool invocation.
  • Ingestion points: User-provided inputs for the paper's title and abstract are used as command-line arguments in SKILL.md.
  • Boundary markers: The instructions lack boundary markers or warnings to prevent the agent from including shell metacharacters or escape sequences in the command interpolation.
  • Capability inventory: The skill utilizes the environment's ability to execute shell commands (python3) to process content from tools/match_venues.py.
  • Sanitization: There are no instructions provided to sanitize, validate, or properly escape the title and abstract strings before they are passed to the shell, creating a vector for command injection.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 24, 2026, 10:29 AM
Security Audit — agent-trust-hub — rt-journal-match