Install Inspector
A teammate just asked you to add a package. You’re one npm install away from running it. Inspect it like a careful engineer, flag what’s suspicious, and decide — before the code runs. Get it wrong and you’ll watch the payload fire.
Loading…
Sample report · illustrative data
Sample result: one install-inspection round
A made-up package, "left-padder-plus", was inspected across its npm page, package.json, files, version diff and provenance, then blocked with three signals flagged.
Round result
- Decision
- Block (correct: the package was malicious)
- Round score
- 130 of 160
- Signals present
- 4
- Caught
- 3
- Missed
- 1
- False flags
- 1
Signals
| Signal | Your flag | Present |
|---|---|---|
| Install script | Flagged | Yes |
| Version anomaly | Flagged | Yes (new files in a patch bump) |
| Obfuscated payload | Flagged | Yes |
| Install-time access | Not flagged | Yes (reads environment variables during install) |
| Typosquat | Flagged | No |
How the round is scored
A correct install or block decision earns 100 points. Each correct signal adds 15, each false flag takes off 10 and each missed signal takes off 5. Installing a malicious package costs 100 points and blocking a safe one costs 40.
Made-up example data. Your own run uses your inputs. Downloads of the full report need Pro.
Frequently asked questions
What is Install Inspector?
An interactive trainer that puts you in a developer’s seat the moment before running npm install. You inspect the package’s npm page, package.json, file list, version diff and provenance, flag the red flags, and decide whether to block or install — then see what the package would actually have done. It’s built to train the one habit that stops supply-chain attacks: inspecting install-time behaviour before trusting the registry.
How do attackers hide malware in an npm package?
The most common vector is a lifecycle script (preinstall / postinstall / install) that runs automatically on npm install, before you ever import or read the code — often an obfuscated blob that harvests environment variables, .npmrc tokens, cloud and SSH keys and exfiltrates them. Others use typosquatting (a name one edit from a popular package), dependency confusion (publishing your internal package name publicly with a huge version), remote second-stage payloads fetched at install, and — the subtle one — abusing a compromised maintainer’s legitimate CI so the malicious release is signed with valid provenance.
Does an install script in package.json mean the package is malware?
No. A lifecycle script is a yellow flag, not a red one. Native modules (sharp, libvips, better-sqlite3, Rust-based tools) legitimately use an install script to download a prebuilt binary from the project’s own release and checksum it. The skill is reading what the script does: fetch-and-verify-a-binary is fine; read-env-and-phone-home is not. The safest default is npm install --ignore-scripts with an allow-list for the few packages that genuinely need a build step.
Can a package signed with provenance still be malicious?
Yes — and this is the trap. Provenance attests where a build came from (which repo, which CI), not whether the code is benign. When a maintainer’s account or CI is compromised, their legitimate pipeline will sign genuinely malicious code with valid provenance. The @subql/common 5.8.3 incident in October 2026 shipped a credential stealer signed by real GitHub Actions. Treat "signed with provenance" as origin evidence, not a trust decision — and diff every version bump, including patches.
Is it free, and do I need an account?
It runs in your browser. It needs Pro, $5/month or $40/year. Your first Cyber War Room incident run is also free with a free account.
More security tools
The tools people use most on PlayCISO.
Catch injected instructions hiding in content your AI agent should only read.
Score your readiness against the first international standard for AI security threats.
Critical CVEs across 21 major vendors, month by month — and what the 2026 surge means.
Build an ordered ruleset, replay labelled packets and catch shadowed and any-to-any rules.
Self-assess all six Functions, from Govern to Recover, and see your weakest one.
Search 172 threat actors across Microsoft, CrowdStrike, Mandiant and community names.