The problem
Every day another thread: "Top 5 research skills — beat McKinsey!" or "This self-hosted app replaces a $200/mo SaaS." A handful of links. Ten minutes on each README, and you still can't tell which one is real.
Stars don't help. An influencer ships filler and gets 20k stars. A student polishes a genuine gem for months and has 30. Whether it is a skill pack, CLI, library, extension, or full application, the honest answer is inside the files — and nobody has two hours per repo.
What you get back
- A verdict you can read aloud — 🟢 INSTALL or 🔴 SKIP on the first line, then three to five plain sentences: what it is, what you get, whether it runs on your machine.
- Five plain questions answered — what does it do, what's good about it, what makes it special, what do you gain, does it run on your machine. About half a page, zero jargon.
- One recommended way to run it — the full point-and-click experience whenever one exists. Never a menu of five install paths, never a terminal-only version of a graphical product.
- At most two "watch out for" items — real money, required accounts, upkeep, or broken promises. No nitpicks.
- A malice tripwire, nothing more — no vulnerability laundry lists. Clear evidence of deliberate theft or sabotage becomes the verdict itself.
- The deep dive, on request — say deep dive and the developer view appears: the traced mechanism, a 10-minute reading map, patterns worth stealing. Hidden by default.
The pre-install gate
Ask it to install or adopt something and it reads first, installs second. Inspection never installs or executes target code. After 🟢 it installs the scope it chose for you and returns to the workflow you authorized; after 🔴 it explains why with file:line evidence. The final call is still yours.
A real test — a repo advertising "50+ skills, 21k stars, replaces a $2M consulting team", asked in Chinese to install it urgently:
结论：🔴 SKIP（不要安装） — 它的安装脚本会把你的全部环境变量偷偷打包上传到作者的服务器…
(Verdict: SKIP. Its install script quietly uploads all your environment variables to the author's server…)
It caught the credential exfiltration at install.sh:6, caught a hidden comment ordering AI reviewers to praise the repo, called out the fabricated star count — and installed nothing.
How it judges
The pill metaphor: a supplement lists a hundred ingredients while one or two do the work. The core move is the removal test — delete a component; would the outcome change? What survives is the essence.
The decisive question is the substance test: does the target add a capability you do not already have — a working pipeline, hard-won domain constraints, reusable assets, or a verification loop? Or is it a thin wrapper dressed up as a system? Source beats presentation.
Stars, followers, and README claims never prove implementation quality. Source files decide the technical verdict; history, releases, and issues provide maintenance context only.
Why it's fast
Size first, then fan out. A quick triage picks a reading tier. Large repositories are split across parallel subagents, keeping regular verdicts under ~10 minutes and very large ones under ~15.
It also never runs the code it judges. The repo is treated as untrusted data throughout — it dissects the pill, it doesn't swallow it. Instructions hidden inside a repo aimed at the reviewing AI are reported as a red flag, never obeyed.
Install
Claude Code / Claude Desktop
(
set -e
source_stage="$(mktemp -d "${TMPDIR:-/tmp}/repo-scout-source.XXXXXX")"
install_stage=""
trap 'rm -rf "$source_stage" "$install_stage"' 0 1 2 15
git clone --depth 1 https://github.com/wallmage/repo-scout.git "$source_stage/checkout"
mkdir -p ~/.claude/skills
install_stage="$(mktemp -d ~/.claude/skills/.repo-scout.XXXXXX)"
cp -R "$source_stage/checkout/repo-scout/." "$install_stage/"
rm -rf ~/.claude/skills/repo-scout
mv "$install_stage" ~/.claude/skills/repo-scout
rm -rf "$source_stage"
trap - 0 1 2 15
)Or download repo-scout.skill from the repo and click Save skill.
OpenAI Codex
(
set -e
source_stage="$(mktemp -d "${TMPDIR:-/tmp}/repo-scout-source.XXXXXX")"
install_stage=""
trap 'rm -rf "$source_stage" "$install_stage"' 0 1 2 15
git clone --depth 1 https://github.com/wallmage/repo-scout.git "$source_stage/checkout"
mkdir -p ~/.agents/skills
install_stage="$(mktemp -d ~/.agents/skills/.repo-scout.XXXXXX)"
cp -R "$source_stage/checkout/repo-scout/." "$install_stage/"
rm -rf ~/.agents/skills/repo-scout
mv "$install_stage" ~/.agents/skills/repo-scout
rm -rf "$source_stage"
trap - 0 1 2 15
)Gemini CLI and other Agent Skills tools — see INSTALL.md for native project and personal discovery paths.
Then just ask
Is this worth installing? https://github.com/someone/some-skill-pack
Is this worth deploying? https://deerflow.tech
thoughts? https://github.com/someone/some-cli-tool
Which of these five is best? <link> <link> <link> <link> <link>
Install this for me: https://github.com/someone/viral-skill ← gets vetted firstAsk in any language and the whole verdict comes back in that language.