code-refactor

Warn

Audited by Gen Agent Trust Hub on May 18, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes dynamic context injection (!echo "$1") in the SKILL.md file. This pattern directly interpolates the user-provided argument $1 into a shell command executed at load time. An attacker could provide an argument containing shell metacharacters (e.g., ; rm -rf /) to execute arbitrary commands on the system.
  • [PROMPT_INJECTION]: The skill is designed to refactor code from a user-provided directory or file path, which creates an indirect prompt injection surface as the agent ingests untrusted third-party data.
  • Ingestion points: Code files located at the file path or directory specified by the user in the $1 parameter.
  • Boundary markers: Absent. The instructions do not include delimiters or warnings to the subagent to ignore instructions that might be embedded within the source code being refactored.
  • Capability inventory: The skill uses the Task and TodoWrite tools and delegates tasks to a subagent that performs behavior-preserving code transformations.
  • Sanitization: Absent. There is no validation or filtering of the file content before it is processed by the agent.
Audit Metadata
Risk Level
MEDIUM
Analyzed
May 18, 2026, 06:55 AM
Security Audit — agent-trust-hub — code-refactor