meta-prompting

Pass

Audited by Gen Agent Trust Hub on May 10, 2026

Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes a local Python script (scripts/copy-to-clipboard.py) which invokes the pbcopy system utility. This is the intended behavior of the skill to provide clipboard integration on macOS systems.
  • [COMMAND_EXECUTION]: The workflow instructions in SKILL.md employ a quoted heredoc (<<'__meta-prompting-skill_EOF__') when passing the generated prompt to the script. This is a robust security measure that prevents the shell from interpreting special characters (like backticks or dollar signs) that might be present in the AI-generated text, effectively mitigating command injection risks.
  • [INDIRECT_PROMPT_INJECTION]: The skill acts on untrusted user-provided draft prompts.
  • Ingestion points: User input is processed and refined into a new prompt.
  • Boundary markers: The skill instructions specify wrapping the output in a unique quoted heredoc sentinel (__meta-prompting-skill_EOF__).
  • Capability inventory: The skill can execute shell commands to run the local clipboard script.
  • Sanitization: The use of quoted heredoc syntax prevents the agent's shell from executing any code contained within the refined prompt string.
Audit Metadata
Risk Level
SAFE
Analyzed
May 10, 2026, 02:36 PM