llm-ops

Pass

Audited by Gen Agent Trust Hub on Mar 10, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill serves as a legitimate technical guide for LLM-Ops. It includes boilerplate code for interacting with the Anthropic API and ChromaDB, following standard implementation patterns.
  • [PROMPT_INJECTION]: The skill implements a RAG (Retrieval-Augmented Generation) pattern which is susceptible to indirect prompt injection if the source documents contain malicious instructions.
  • Ingestion points: Functions rag_query and evaluate_response in SKILL.md interpolate external data and model outputs directly into the final prompt.
  • Boundary markers: The prompts use simple textual labels (e.g., 'Contexto:', 'PERGUNTA:', 'RESPOSTA ATUAL:') to separate data but do not employ robust XML-style delimiters or explicit instructions to ignore embedded commands.
  • Capability inventory: The skill utilizes the anthropic Python client to execute LLM queries based on the constructed prompts.
  • Sanitization: There is no evidence of input sanitization or validation in the provided code snippets.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 10, 2026, 01:51 PM