Show HN: Second opinion – A skill to query different models
A developer created a skill called "Second Opinion" that allows querying different AI models (Claude or opencode-supported models) mid-session to cross-examine suggestions. The tool compiles relevant context, consults a second model, and supports follow-up questions. Currently supports Claude-to-Claude, Claude-to-opencode, and opencode-to-Claude invocations.
Background
- This is a "skill" (a custom tool or plugin) for AI coding assistants, shared on Hacker News via "Show HN" — a section where makers show projects they built.
- The skill lets you send the current AI model's suggestion to a *different* AI model for a "second opinion" during a single work session. Currently supports switching between Claude (Anthropic) and Opencode (an open-source coding assistant compatible with many AI backends).
- The idea: if Claude gives you advice, you can immediately ask another model (e.g., Opencode or a different Claude instance) to critique or verify it — like asking a colleague to check your work. The skill compiles the relevant context automatically so you don't have to re-explain.
- "OpenCode" refers to an open-source terminal-based coding assistant (opencode-ai) that supports models from OpenAI, Anthropic, Google, and others. "Claude" is Anthropic's large language model.
- This matters because a common frustration with AI coding helpers is over-reliance on one model's blind spots; cross-examination aims to catch errors or alternative approaches without manual copy-pasting.