Sends named code files to the OpenAI API for an independent adversarial review.
npx skills add https://github.com/timikalo7/Execute --skill second-opinion-timikalo7
Please help me install this Agent Skill. Skill: second-opinion Source: https://github.com/timikalo7/Execute/tree/main/.claude/skills/second-opinion Command: npx skills add https://github.com/timikalo7/Execute --skill second-opinion-timikalo7
--ask flag lets you pose a specific concern (e.g., a suspected edge case in call-site resolution) alongside the general review. - Explicit failure diagnosis: Detects quota and billing errors and states plainly that a ChatGPT/Codex subscription does not include API credit, so the failure is not mistaken for a bug. - Use Case: After a build passes every internal quality gate, run the script on the core module with a pointed question to get a dissenting review before shipping, then verify each finding against the code before acting on it. ## Quick Start Set OPENAI_API_KEY in the environment and run node review.mjs with the files to review, optionally adding --ask with a specific question.High-intent search queries and answers about installing and using this skill.
Run the review.mjs script with the files to review as arguments and OPENAI_API_KEY set in the environment. It sends only those files to the OpenAI API with an adversarial prompt and prints findings with file, line, and failing input.
Pass the --ask flag followed by your question, for example asking whether call-site resolution has a silently wrong case. The question is appended to the review prompt so the model addresses it directly alongside general findings.
A ChatGPT or Codex subscription does not include API credit; they are separate billing meters. Add a balance at platform.openai.com under Billing, and the script explicitly names this quota failure when it occurs.
Only the files named on the command line are sent, and the API key is read from the environment and never written anywhere. Read files before sending and never include personal data without the owner's explicit decision.
Treat the output as one more opinion, not a verdict. A different model is confidently wrong in different places, so verify every finding against the actual code before acting on it.
timequityAutomates verification of tests, builds, linting, and security checks before completion.
pedrohcgsVerify review findings against source material before applying fixes.
cpfcoachingEvaluate critical development decisions through fresh-context adversarial reviews.
m-xlseaBridge Codex CLI with client apps for single-turn and multi-turn interactions.