Security scorecard
Good
Scanned 2026-11-09 at 4d51724
Score
6.2
Out of 10
Passing
7/12
Checks at 7+
Weak
4
Checks under 4
Stars
143,749
32,818 forks
Last push
today
2026-11-18
Licence
MIT
JavaScript
Action plan
The weakest checks, highest risk first. Maintainers can use this as a to-do list; users can use it to judge what to watch.
1. Token-Permissions
High0/10Set `permissions: read-all` (or `contents: read`) at the top of each workflow and grant write scopes only to the specific jobs that need them.
2. Binary-Artifacts
High0/10Remove committed binaries and build them from source in CI. If one is unavoidable, document it and verify its checksum during the build.
3. SAST
Medium0/10Enable a SAST tool on every pull request, such as GitHub code scanning with CodeQL.
4. Pinned-Dependencies
Medium4/10Pin GitHub Actions to full commit SHAs, Docker images to digests, and use lockfiles with hashes. Let your update tool bump the pins.
5. CII-Best-Practices
Low0/10Complete the self-assessment at bestpractices.dev and display the badge in the README.
14 checks
Open a row for what the check looks at, the raw result and the recommended fix. N/A means the check does not apply (for example, no releases or no workflows).
What it checks
Looks for CI workflows that run untrusted pull-request code with write access, or that paste user-controlled text (titles, branch names) straight into shell steps.
Such patterns let an outside contributor run code with the repository's secrets, which can mean a full repository or release compromise.
Result
no dangerous workflow patterns detected
Keep it that way
Avoid pull_request_target / workflow_run jobs that check out the PR head. Pass event fields through environment variables instead of inline ${{ }} expressions in run steps.
What it checks
Reviews the GITHUB_TOKEN permissions declared in workflows.
Over-privileged tokens turn any compromised build step into write access to the repository, packages or releases.
Result
detected GitHub workflow tokens with excessive permissions
Recommended fix
Set `permissions: read-all` (or `contents: read`) at the top of each workflow and grant write scopes only to the specific jobs that need them.
What it checks
Detects compiled executables (jars, .exe, .pyc, .so and similar) committed to the source tree.
Binaries cannot be reviewed like source, can silently drift from the code, and are an easy hiding place for malicious payloads.
Result
binaries present in source code
Recommended fix
Remove committed binaries and build them from source in CI. If one is unavoidable, document it and verify its checksum during the build.
What it checks
Measures how many recent changes were reviewed by someone other than the author before merging.
Review is the main defence against both honest mistakes and deliberately injected code.
Result
Found 28/30 approved changesets -- score normalized to 9
Keep it that way
Merge through pull requests only and require an approval from a second maintainer, including for maintainers' own changes.
What it checks
Counts commits and issue activity over the last 90 days and flags archived or very new repositories.
An unmaintained project is unlikely to patch new vulnerabilities, so problems stay open indefinitely.
Result
30 commit(s) and 3 issue activity found in the last 90 days -- score normalized to 10
Keep it that way
Keep a visible cadence of fixes and triage. If the project is finished or abandoned, say so in the README or archive it so users can plan a replacement.
What it checks
Inspects the rules on the default and release branches: required reviews, status checks, blocked force pushes and deletions, and whether admins are covered.
Without protection, one compromised or careless account can push directly to the code everyone installs.
Result
internal error: error during branchesHandler.setup: internal error: githubv4.Query: Resource not accessible by integration
When it applies
Enable a branch protection rule or ruleset: require at least one (ideally two) approving reviews, require status checks, dismiss stale approvals, block force pushes and include administrators.
What it checks
Checks whether recent release assets come with signatures or provenance attestations (for example Sigstore bundles or SLSA provenance).
Unsigned releases give users no way to tell a genuine artefact from a tampered one.
Result
no releases found
When it applies
Sign release artefacts in CI (cosign / Sigstore) and publish SLSA provenance alongside them.
What it checks
Checks whether static analysis (for example CodeQL, Semgrep or SonarCloud) runs on changes.
Static analysis catches common vulnerability classes before they are merged.
Result
SAST tool is not run on all commits -- score normalized to 0
Recommended fix
Enable a SAST tool on every pull request, such as GitHub code scanning with CodeQL.
What it checks
What it checks
Checks whether the project holds an OpenSSF Best Practices badge, and at which level.
The badge shows the maintainers have worked through a recognised list of secure development practices.
Result
no effort to earn an OpenSSF best practices badge detected
Recommended fix
Complete the self-assessment at bestpractices.dev and display the badge in the README.
What it checks
Checks for a recognisable licence file in the repository.
Without a clear licence, users and security reviewers cannot legally rely on or audit the code.
Next steps
Scores come from weekly automated scans and repository details from the public GitHub API. Results are cached for 24 hours (last fetched 2026-11-18 18:43 UTC). Automated checks are a signal, not a security audit.
aidir3 Engineering
Not sure where to start? Pick your problem on the developer hub or browse all engineering tools.
Checks whether build dependencies are pinned to exact versions or hashes: GitHub Actions, Docker base images, and install commands in scripts.
Floating references (a tag, `latest`, curl | bash) can change underneath you, so a compromised upstream lands in your build silently.
Result
dependency not pinned by hash detected -- score normalized to 4
Recommended fix
Pin GitHub Actions to full commit SHAs, Docker images to digests, and use lockfiles with hashes. Let your update tool bump the pins.
What it checks
Looks for a CI workflow that publishes the project as a package to a registry.
Packaged releases make it simple for users to receive and apply security updates.
Result
packaging workflow detected
Keep it that way
Publish from a CI workflow to the ecosystem's registry (npm, PyPI, crates.io, container registry) rather than from a laptop.
What it checks
Checks for a SECURITY.md that explains how to report vulnerabilities privately.
Without one, researchers may disclose a flaw publicly or not at all.
Result
security policy file detected
Keep it that way
Add SECURITY.md with a private contact or GitHub private vulnerability reporting, supported versions and expected response times.
What it checks
Detects fuzz testing, for example OSS-Fuzz or ClusterFuzzLite integration or native fuzz tests in Go, Rust, Python and others.
Fuzzing finds memory-safety bugs and crashes on unexpected input that ordinary tests miss.
Result
project is fuzzed
Keep it that way
Add fuzz targets for parsers and input-handling code and run them continuously (OSS-Fuzz for widely used projects, ClusterFuzzLite in CI otherwise).
Result
license file detected
Keep it that way
Add a LICENSE file at the repository root with an OSI-approved licence.