jia-ethan/github-solution-research

GitHub Solution Research is a skill for the Codex AI agent that turns concrete engineering problems into structured GitHub research. When you hit a technical blocker, the skill searches public repositories, issues, pull requests, discussions, code examples, and release notes to find how others have already handled similar problems. The workflow follows a set sequence: you describe the problem with its error text, version numbers, and constraints, the skill uses the GitHub CLI (gh) to search, it ranks results by how well they match the problem, it reads the strongest matches in depth, and it extracts patterns that can be applied locally. The output includes candidate repositories with stars, forks, license, and activity level, along with the specific issues or PRs that contain relevant evidence, a recommendation for what to reuse versus what to adapt, and a verification step such as a test command. Subagents can be used for research that spans multiple ecosystems or query types, but only conditionally. The main controller keeps responsibility for deduplication, ranking, and independently verifying any claims the subagents return. The skill is installed by cloning the repository into the Codex skills directory and requires the GitHub CLI to be installed and authenticated. It works with any AI model. The README is clear about what the skill does not do: it is not an automatic code fixer, not a vulnerability scanner, and does not guarantee GitHub contains an answer. It should not be used to access private repositories, copy credentials, or handle production data. It is licensed under MIT. The README is bilingual in Chinese and English.