gh-scoutlisted

Deep GitHub reconnaissance before planning or building a feature. Given a description of what you're about to build, finds existing repos that (a) already do it — so you can adopt instead of rebuild — and (b) are complementary building blocks you could depend on. Use before writing an implementation plan, when scoping a new feature/library/tool, when evaluating "should we build or adopt this", or when the user says "scout github", "any libraries for this", "has someone built this", or invokes /gh-scout.
kosmoperion/gh-scout · ★ 1 · AI & Automation · score 72

Install: claude install-skill kosmoperion/gh-scout

# gh-scout Reconnaissance, not a plain search. GitHub's native search is keyword-matched and shallow. The value here is the **LLM layer on both ends** of it: you expand one intent into many search angles, then read the top candidates' READMEs and judge real relevance. A run costs a minute and can save days of rebuilding something that already exists as a maintained library. ## When to run this - **Right before planning or building** anything non-trivial — a new feature, service, library, CLI, or integration. - When the question "should we build this or adopt something?" is live. - When the user explicitly asks to scout, or types `/gh-scout`. If the task is a trivial change to existing code, skip it — this is for greenfield capability, not edits. ## Context hygiene: run it in a subagent This process pulls dozens of repo records and ~10 READMEs — a lot of noisy tokens right when you want a clean context to plan in. **Prefer delegating the whole run to a subagent** (the `Explore` or `general-purpose` agent) and having it return only the final digest. Give the subagent this SKILL.md's steps and the intent; let it hand back the ranked report. Run inline only for a quick, narrow lookup. ## The pipeline ### 1. Distill intent Write one or two sentences: *what capability* is being built, and in *what ecosystem* (language, framework, domain). This is what you expand from. ### 2. Expand into 6–10 search queries This is the step that makes results good. Brainstorm diverse angl