Launchrail's job is writing files into other people's repositories, so its security bar is concrete: nothing it does may exceed what the ownership model and manifest declare.

Report vulnerabilities privately via GitHub private vulnerability reporting ("Report a vulnerability" on the repo's Security tab). Please do not open public issues or pull requests for security problems.

Include what you can: affected command or module, a reproduction (ideally against a scratch repository), and impact. You can expect an acknowledgement within 7 days. Coordinated disclosure is appreciated; we'll credit reporters in release notes unless you prefer otherwise.

Beyond the usual (dependency compromise, code execution), Launchrail-specific classes we care about most:

• Ownership-model escapes — any write that overwrites a seeded or project-owned file, or touches paths outside the consuming repository (path traversal via templates, module names, or manifest values).
• Checksum/dry-run bypass — a code path that writes without dry-run support, or replaces a locally modified managed file without detecting the conflict.
• Migration damage — a failed migration that leaves a repository unrecoverable.
• Template or manifest injection — untrusted content in .launchrail.yml, lockfile, or upstream templates causing writes or command execution the user didn't ask for.
• Secret leakage — anything that causes secrets to be written into the manifest, lockfile, templates, or evidence bundles.

Skills and workflow scripts installed into consuming projects (.claude/, scripts/) are instructions executed by coding agents and Node; treat injection into those artifacts as in scope too.

Pre-release: only the latest published version (and master) receives security fixes. This section will be updated with a version table once 1.0 ships.