Install Check — Security skill for Claude Code
Audit a skill file, plugin or MCP connector before you install it.
How to install Install Check
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open ToolMonsters/install-check and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What Install Check does
Audit a skill file, plugin or MCP connector before you install it. Five checks, one verdict.
Alternatives in Security
- Install — Install the Damage Control security hooks system 456 ★
- Unifly Audit — Security audit of UniFi configuration 248 ★
- Marketplace — Security-audited skills for Claude, Codex & Claude Code 247 ★
README
install-check
**Read the file before you run it.**
People install skill files, plugins and MCP connectors the way they used to install browser extensions: they don't read them. A file that arrives as a link, a zip or a paste has the same reach as software you install, and almost nobody opens it first.
This skill opens it for you.
What it does
Give it a `SKILL.md`, a plugin folder, an MCP config, or the install instructions someone posted. It reads the actual file and answers five questions:
- Where do the links go? Every URL and package name, and whether each one is the official source or a lookalike.
- What does it ask you to run? Every shell command,
curl | bash,npx, or "paste this" instruction, in plain words. - What can it reach? Which tools, accounts and files, compared to the job it claims to do.
- What leaves the room? Anything that sends, posts, uploads or emails, and whether the file says so openly.
- Who wrote it? Author, date, version, licence, repository, and what is missing.
Then a verdict: **INSTALL**, **READ FIRST**, or **DO NOT INSTALL**, with the exact lines to look at.
What it does not do
It is not an antivirus. It does not scan binaries and it does not check domains against a threat database. It reads text and flags what a careful person would flag if they took the time to read, which nobody does.
That is the whole point.
Prompt injection
A skill file is a file of instructions, so a malicious one can contain text aimed at the model reading it. This skill treats every audited file as data to be quoted, never as instructions to follow, and reports any attempt to direct its behaviour as a finding.
Install
Download `SKILL.md` and add it to your assistant the same way you add any other skill.
Then, before you trust it: run its own five checks on it. It has an author, a version, a licence and a public history, so all five are answerable.
Built by Yonathan Cohen · [Tool Monsters](https://to
Related Skills
Skills Repo
Five self-contained agent skills for Claude — interview before building, restructure brain dumps, coach beginn
Security Skills
Two skills for reviewing what you are about to trust: an AppSec/CloudSec/CI-CD review pass, and an audit that
IntentProbe
Activation-probe security scanner for AI agent tooling. Reads a model's internal activations to detect poisone
n8n Google Meta Ads
n8n workflows: AI agent on your Google Ads + Meta Ads via the Ryze MCP (https://connector.get-ryze.ai/mcp), Ga
Fable Skills
Five skills Fable wrote before it left: the security sweep, the setup, the build planner, the honest advisor,
Audit Note
Re-run the two-layer faithfulness audit on a single Synapse note. Useful after manual edits, after a prompt ch
Related Agents
Pin Auditor
Audits validator checks for VACUITY — a pin that reports green against the mutation it was written to catch. R
Issue Claim Auditor
PRFlow's implement-phase Issue-Claim Audit agent. Runs Phase 1.6's specification-projection check and targeted
Gate Auditor
Adversarially audit a test suite or release gate by trying to make every green check go red. Use before trusti