scanrepo

Scan any GitHub or Bitbucket repo for malware, credential stealers and crypto scams — before you clone it.

npx scanrepo github.com/owner/repo

No install, no signup, no clone. One command, a verdict in seconds.

Why

A "recruiter" sends a take-home assignment. A trending repo promises a trading bot. Someone just needs you to run it locally real quick. Then npm install fires a postinstall that reads your browser profiles, SSH keys and wallet files — or a .vscode/tasks.json runs the moment you open the folder.

This is the check you run first, in the two seconds before you type git clone.

Usage

npx scanrepo github.com/owner/repo          # human-readable report
npx scanrepo bitbucket.org/ws/repo --json   # raw JSON on stdout
npx scanrepo <url> --token ghp_...          # GitHub PAT — full coverage on big repos
npx scanrepo <url> --no-publish             # don't save the result to scanrepo.dev
npx scanrepo <url> --no-color               # plain output for logs

Prefer it on your PATH: npm i -g scanrepo. Requires Node 18+.

Example

Risk Score: 100/100  MALICIOUS
────────────────────────────────────────────────
  ⚠  Created 3 days ago
  ℹ  Zero stars, forks, and issues. Common for throwaway scam repositories.

  Code Execution      █░░░░░░░░░░░░░░░  2/25
  File System Access  ███████████░░░░░  14/20
  Obfuscation         ████████████████  15/15
  Supply Chain        ██░░░░░░░░░░░░░░  2/15

Top findings:
  clipboard-read                src/dev/tauri-mock/clipboard-manager.ts
  drainer-secret-collection
  long-lines                    src/dev/tauri-mock/fs.ts
  high-entropy-strings          src/App.tsx

  github  owner/repo  ·  102/173 files scanned  ·  master

  ❌ DO NOT RUN

Exit codes

Script-friendly — the exit code fails your step for you, no parsing needed.

| Code | Meaning | | --- | --- | | 0 | safe / low — nothing matched | | 1 | suspicious, an inconclusive scan, or a scan error | | 2 | dangerous / malicious — do not run it |

An inconclusive result means too few of the repo's files could be read to stand behind a clean verdict. It prints INCONCLUSIVE with the coverage percentage and exits 1 rather than handing you a green light the scan never earned.

In CI:

- name: Scan before running third-party code
  run: npx scanrepo github.com/${{ github.repository }}

What it detects

A static rules engine scores the repo across six categories:

| Category | Examples | | --- | --- | | Code execution | eval, Function(), child_process, vm abuse | | Network exfiltration | fetch-then-eval, hardcoded IPs, suspicious domains | | Filesystem access | browser profile theft, wallet files, SSH credential reads | | Obfuscation | base64 payloads, hex blobs, string-decode chains, entropy | | Supply chain | postinstall scripts, known-malicious dependencies | | Web vulnerabilities | XSS sinks, open redirects |

On top of the rules: entropy analysis, an import/reachability graph, repo age and popularity signals, and a curated IoC list of known-malicious repos and accounts. Scores roll up to 0–100 and a verdict — safe · low · suspicious · dangerous · malicious.

Limitations

Read the score as evidence found, not as a safety guarantee.

• Static analysis only. Heavily obfuscated malware can evade pattern matching. safe means no known pattern matched.
• Public repos only. Private repos are not supported.
• Rate limits shape coverage. Anonymous GitHub/Bitbucket API limits can cut a large scan short — pass --token (or set GITHUB_TOKEN) for full coverage.
• Results are published to scanrepo.dev so others benefit from the scan. Opt out per run with --no-publish.

Related

• Web app — scanrepo.dev
• MCP server — scanrepo-mcp, the same scans for your AI agent, before it clones anything
• IoC list — rubenmarcus/malicious-repositories

MIT © Ruben Marcus