agent-orchestration-multi-agent-optimize

Pass

Audited by Gen Agent Trust Hub on May 17, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill concludes with Target Optimization: $ARGUMENTS. This pattern interpolates untrusted user input directly into the agent's prompt context without any boundary markers or sanitization, allowing for potential override of the agent's core instructions and behavior.
  • [PROMPT_INJECTION]: The skill lacks explicit boundary markers (such as XML tags, triple backticks, or unique delimiters) to separate system instructions from the data provided in the $ARGUMENTS variable, creating a surface for indirect prompt injection.
  • Ingestion points: The $ARGUMENTS parameter in SKILL.md acts as a direct entry point for user-controlled data.
  • Boundary markers: No isolation techniques or delimiters are employed to wrap the interpolated content.
  • Capability inventory: The skill's stated purpose involves performance profiling and optimization, which typically requires capabilities such as file system access, network operations, or shell execution to analyze system components (databases, applications, frontends).
  • Sanitization: There are no instructions or logic provided to validate, escape, or sanitize the input before it is added to the agent's context.
Audit Metadata
Risk Level
SAFE
Analyzed
May 17, 2026, 10:32 PM
Security Audit — agent-trust-hub — agent-orchestration-multi-agent-optimize