openrouter-performance-tuning

Pass

Audited by Gen Agent Trust Hub on Sep 9, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to process user-provided text prompts and messages through the OpenRouter API. This creates a standard surface for indirect prompt injection if the ingested data contains malicious instructions intended to influence the agent's behavior.
  • Ingestion points: Multiple functions across SKILL.md and reference files (e.g., benchmark_model, stream_completion, parallel_completions, efficient_batch, monitored_chat) accept raw string input from the user or agent context.
  • Boundary markers: The provided implementation examples do not use explicit delimiters or "ignore embedded instructions" warnings for the prompt content.
  • Capability inventory: The skill utilizes network access to communicate with the OpenRouter API and uses the Bash(python3:*) tool to execute measurement and optimization scripts.
  • Sanitization: No specific sanitization or filtering of the input content is implemented in the provided snippets, as is typical for general LLM wrapper tools.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 9, 2026, 03:46 AM
Security Audit — agent-trust-hub — openrouter-performance-tuning