Vet Before You Install: ten checks, a sandbox recipe and a skill for third-party agent code
by PaperChaseWebb
A Markdown pack for deciding whether a skill, plugin, mod, MCP server or repo is safe to let into your agent setup: read it first, pin and vendor it, trial it in a sandbox with a planted fake key, then write the decision down.
- Price
- $39
- Difficulty
- Intermediate
- Setup
- 15 min
- Format
- .md file
vet-before-you-install.md · 15 KB · instant download after checkout
Want this built for you instead? We can build it for you.
What you end up with
Nothing third-party runs on your machine until it has been read, pinned, trialed on disposable code and given a dated decision.
What you get
- Why star counts are not vetting
- The ten checks
- A read-first walkthrough with a dangerous token scan
- Pin and vendor commands
- A sandbox recipe for containers, Linux and macOS with a planted key test
- 15 prompts
- A drop-in vet-before-install SKILL.md
- Vetting card, adoption log and reject note templates
- Special cases: in-process mods, share-alike licenses, reference implementations
- A credited source list
Who it is for
People who see "install these 40 skills" posts and want a routine that keeps unknown code away from real projects and real secrets.
Works with
Solid chips: the agent is named in this listing's own tools list. Dashed chip: Plain Markdown, so it also pastes into other agents such as Grok Bot, Hermes, OpenClaw or Muse. We list those as compatible by format only, not as tested.
Tools needed
Cost to run: Free. Docker is optional.
Time to result: First vetting card in one sitting
Sample prompts, free to use
In ~/vetting/REPO, list every file that executes or that a skill instructs an agent to run. Explain each in plain words. Quote any line that touches the network, the shell, the environment or files outside its folder. Do not run anything.
Validators and helper scripts are where surprises hide.
Free preview of the file
# Vet Before You Install A PaperChaseWebb pack, version 1.0. A vetting routine, prompts, a sandbox recipe and an agent skill for deciding whether a skill, plugin, mod, MCP server or repository is safe to let into your agent setup. Written for Claude Code, Codex, Cursor and Grok Bot style agents on macOS or Linux. This pack is not a security audit and does not certify anything as safe. It is the routine we use before any third-party code touches a real machine, rebuilt in our own words from public posts, public repositories and our own vetting notes. Sources are credited in `10-sources.md`. ## Files | File | What it is for | |---|---| | 01-why-stars-are-not-vetting.md | What a star count measures, what it does not, and the question to ask instead | | 02-the-ten-checks.md | The ten checks every candidate passes before it runs anywhere you care about | | 03-read-it-first.md | How to have an agent read the whole skill or repo, line by line, without running it | | 04-pin-and-vendor.md | Pinning a commit, verifying it, and keeping your own copy instead of a one-line remote install | | 05-sandbox-trial.md | A disposable trial recipe: no network, read-only host, allowlisted environment, a planted fake key | | 06-prompts.md | 15 prompts, from intake to the adopt or reject decision | | 07-vet-skill.md | A drop-in SKILL.md that makes an agent refuse to install anything it has not vetted | | 08-templates.md | VETTING-CARD.md, ADOPTION-LOG.md and a reject note | | 09-special-cases.md | In-process mods, reference implementations, share-alike licenses, near-identical names | | 10-sources.md | Every source we learned from, linked and credited |
Worth knowing
- This is a routine, not a security audit or a certification.
- A sandbox lowers risk; it does not remove it.
Questions
Do I need Docker?
No. The pack includes an env -i and prlimit recipe for Linux and a separate user or VM approach for macOS.
Does it recommend specific skills?
No. Public repos are named only as examples of things worth vetting, with their licenses as stated by each repo.
Details
Version 1.0 · Updated Oct 26, 2026 · Markdown version of this page
License: Personal and commercial use inside your own work. Do not resell or republish this pack. Download terms
Related
Vet Before You Install: the ten checks (free sample)
The ten checks to run before any third-party skill, plugin or repo touches a real machine, plus one read-first prompt.
Pre-launch security prompts for vibe-coded apps: Supabase RLS, leaked keys and open doors
14 prompts that make your coding agent find exposed keys, tables without row level security, open buckets and unverified webhooks in your own app, then fix them one commit at a time.
Check the Claim (expanded edition): a method, prompts and a skill for checking AI claims
A Markdown pack that makes your agent check income, tool, popularity, quote and funding claims before they reach a brief, a purchase or a post: arithmetic first, primary sources second, one honest verdict per claim.