patchline.dev

Security

Trusted with read access, designed like it

Patchline reads customer source code, which makes security posture the product. This page states what we do in the same terms as our repository's SECURITY.md — each claim below is cross-checked against the implementation and its tests.

What happens to your code

GitHub App permissions, and why each exists

PermissionLevelWhy
Repository contentsread-onlyIndex which provider SDKs you use and where they are called
IssueswriteDeliver blast-radius reports as labeled issues on affected repos
Pull requestswriteOpen migration PRs for repos that opt in — never auto-merged
MetadatareadBasic repository info: name, default branch

That is the complete list. No admin scopes, no workflow scope, no organization-level access.

Sandboxing the migration agent

Generated patches are validated by running the target repo's own test suite in an isolated environment — user namespaces where available, otherwise docker run --network=none. Network isolation is verified by an automated test, and the isolation level achieved is recorded in every confidence score: an un-isolated run is penalized and can never silently upgrade to a PR.

Webhooks, secrets, and supply chain

Reporting a vulnerability

Email rvasqz86@gmail.com with subject [patchline security]. You will get an acknowledgment within two business days. Please do not open public issues for security reports; we credit reporters in release notes unless you prefer otherwise.