repo-troubleshooting-guidelisted
Install: claude install-skill scoobydrew83/skills
# Repo Troubleshooting Guide
Turn a repository into a structured, professional troubleshooting guide. The guide is
grounded in two sources: (1) the **detected stack** (languages, frameworks, services,
infra) and (2) **mined evidence** from the actual code (raised exceptions, human-readable
error strings, env vars, ports, Docker services, CI, and TODO/FIXME markers). Generic
advice is a fallback; repo-specific evidence always wins.
## What you produce
A single Markdown file (default `TROUBLESHOOTING.md` at the repo root) with: a quick
setup checklist, issue sections organized by subsystem, recovery procedures, a known-gaps
section from TODO markers, and a "still stuck" section. Clean professional tone — no emoji,
no motivational filler. See `references/output-template.md` for the exact structure.
## Workflow
Follow these phases in order. Each phase has a clear endpoint so you can checkpoint.
### Phase 1 — Locate the repo
Determine the repository path. If the user uploaded or cloned a repo, use that path. If
they named a public repo, clone it (shallow): `git clone --depth 1 <url> /tmp/<name>`.
If the path is ambiguous, ask which directory is the repo root before proceeding.
### Phase 2 — Scan for evidence
Run the scanner to produce a structured inventory. It is stdlib-only Python and writes
JSON describing the stack and failure-mode signals:
```bash
python scripts/scan_repo.py <repo-path> --out /tmp/evidence.json
```
Read `/tmp/evidence.json`. Key fields:
- `languages`