batch-process

Warn

Audited by Gen Agent Trust Hub on Jun 16, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes shell utilities like ls and find to resolve file lists from a user-provided input_glob. This pattern is vulnerable to command injection if the glob pattern or resulting filenames contain shell metacharacters.
  • [REMOTE_CODE_EXECUTION]: The skill features a custom operation that enables the execution of arbitrary Script-Fu expressions. This creates a dynamic execution environment where user-supplied logic is executed by the GIMP process.
  • [PROMPT_INJECTION]: The skill presents an indirect prompt injection surface by interpolating untrusted data into executable scripts.
  • Ingestion points: File paths and names from the local system, along with user-provided operation arguments in SKILL.md.
  • Boundary markers: None; there are no delimiters or instructions to treat data as non-executable.
  • Capability inventory: The skill can enumerate the filesystem (ls, find) and execute scripts via run-script-fu in SKILL.md.
  • Sanitization: None; there is no logic described for escaping or validating input before script composition.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jun 16, 2026, 11:26 PM
Security Audit — agent-trust-hub — batch-process