This is a Claude Code plugin that scaffolds repo-hygiene artifacts and installs a committed, dependency-free lint plus a CI workflow into target repositories. It processes no user data and runs no service. Its security surface is what it writes into other repos: the payload lint, the CI workflow, the templates, and the harness engines that generate them.

In scope for this policy: vulnerabilities in the toolkit itself, for example a payload or template that would execute unexpected code in a target repo's CI, a harness engine that writes outside the paths it declares or follows a hostile path out of the target directory, a code path that silently overwrites operator content the design says must be refused, or an injection route from a sensed repo's content into what the scaffolder writes.

Please report privately. Do not open a public issue for a security report.

• Preferred: email dev@runverdict.com.
• Collaborators: GitHub private vulnerability reporting, from the Security tab of runverdict/repo-standard-toolkit → Report a vulnerability.

Include the affected file/engine, a minimal reproduction (the engines are pure and dependency-free, so a short node harness/<engine>.mjs … invocation or a failing acceptance/test-*.mjs case is ideal), the impact, and any fix you have in mind.

• Acknowledgement within 3 business days.
• An assessment and, for a confirmed issue, a remediation plan within 10 business days, plus (where the property is determinizable) a new standing acceptance/test-*.mjs that fails the build if the issue ever recurs.
• Credit in the release notes if you would like it; coordinated disclosure once a fix has landed.

This is a community project maintained on a best-effort basis; these are targets, not a contractual SLA.

• The hygiene of a repo the plugin was not run against, and drift a target repo introduces after removing or editing the installed lint. The enforcement is theirs once installed.
• The behavior of the Claude model driving the skill. The design assumes the agent can err, which is why everything that must hold is enforced by the committed lint in CI, not by the agent.

The plugin ships from main; fixes land there and go out in the next version bump. There is no back-port of security fixes to older tags; update to the latest main or release.