clipboard

Pass

Audited by Gen Agent Trust Hub on May 9, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes shell commands such as pbcopy, cat, and rm to manage the system clipboard and temporary files.
  • [COMMAND_EXECUTION]: It uses swift -e to execute a script that interfaces with the macOS AppKit framework for rich text support.
  • [PROMPT_INJECTION]: The skill has an indirect prompt injection surface because it processes potentially untrusted text from the conversation and interpolates it into shell commands.
  • Ingestion points: Conversation text blocks are used as input for clipboard commands (SKILL.md).
  • Boundary markers: The skill explicitly instructs the agent to use single-quoted heredoc delimiters (e.g., 'CLIPBOARD'), which prevents the shell from expanding variables or executing commands within the content.
  • Capability inventory: Local command execution, file system writes to /tmp, and macOS system clipboard access (SKILL.md).
  • Sanitization: Relies on the host shell's single-quoted heredoc mechanism to treat user content as literal data.
Audit Metadata
Risk Level
SAFE
Analyzed
May 9, 2026, 10:57 PM
Security Audit — agent-trust-hub — clipboard