multi-model-meta-analysis

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill explicitly instructs the agent to read source files and "quote relevant code" as evidence, which can require reproducing hardcoded API keys, tokens, or passwords verbatim from the codebase — creating an exfiltration risk.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.80). The skill explicitly ingests user-pasted outputs from external models ("User pastes feedback/analysis from multiple LLMs (Claude, GPT, Gemini, etc.)") which are arbitrary, untrusted third-party content the agent will parse and interpret as part of its workflow, enabling potential indirect prompt injection.