| name | second-opinion |
| description | Consult peer LLM for alternative perspectives when you need second opinion, peer review, or fresh take on technical decisions; useful for validating approaches or exploring alternative solutions |
| allowed-tools | Read, Bash |
Second Opinion Skill
Use this skill when you need an alternative perspective on technical decisions, architecture choices, or implementation approaches.
Usage
Invoke the consultation script using the base directory shown above when this skill loads:
bash <base_directory>/scripts/consult.sh --message "your question" --files path1,path2,path3
Important:
- Replace
<base_directory> with the actual "Base directory for this skill" path shown above
- The script will look for
OPENAI_API_KEY in this order: .env.local, .env, system environment variable
- Default timeout is 30 minutes; match this in your Bash tool timeout parameter (1800000ms)
- Do NOT run in background—synchronous execution avoids polling overhead
Note on GPT-5 models: GPT-5 was released on August 7, 2025. Available variants: gpt-5, gpt-5-mini, gpt-5-nano, gpt-5-pro (thinking model). Default is gpt-5-pro-2025-10-06. Override via SECOND_OPINION_MODEL env var.
Context Responsibility
Critical: You must pass ALL relevant context. The peer consultant has no access to this codebase, conversation history, or project context beyond what you provide.
Include:
- Complete problem description
- Relevant code files (via --files)
- Constraints, requirements, decisions made
- Specific questions or areas for review
Interpretation Authority
You retain final judgment. The peer consultant provides alternative viewpoints, but you must:
- Evaluate responses against conversation context
- Reject/question suggestions that conflict with established requirements
- Apply your understanding of the full project context
- Make final decisions based on complete information
The peer opinion is input, not directive.
