See what your ECR lifecycle policy will delete before it breaks a deploy
One multi-arch push creates several rows in ECR: an image index, one image per platform, and provenance or SBOM attestations. Only the index has a tag. A rule that expires untagged images can delete the per-platform images of an image you still run, and the next pull fails. Connect an AWS account with a narrow read-only role and we show every repository as images grouped the way they were built, dry-run your lifecycle policies ourselves (no write call to AWS), and warn when a rule would remove something a kept image or a running service still needs.
What it does
- Grouped image view: index, then the image for each platform (os/arch), then attestations, signatures and SBOMs. Shows tags, sizes, push and last-pull dates.
- Docker labels from each image's config (version, git commit, source), which the ECR console does not show.
- Untagged but needed: marks untagged rows that a tagged image depends on, so nobody deletes them by hand.
- Lifecycle policy dry-run: runs your rules (tag status, prefixes, wildcard patterns, image count, days since push or pull, archive) and lists what they would expire or archive today and in the next 7 days.
- Loud warnings when an expiring row is part of a kept tagged image, its signature or provenance, or a digest that an ECS service, a running task or a Lambda function uses.
- A suggested safer policy in JSON, checked by running the same dry-run on it.
- Storage report: storage per repository and per kind (images vs attestations and signatures), the largest repositories, and how much the policy would free.
- Daily scan with Slack or email alerts when a new risky outcome appears.
Read-only: the role can describe repositories and images, read manifests, lifecycle policies and image config files, and list what ECS and Lambda run. It cannot push, delete, tag or change a policy. We store digests, tags, sizes, dates and labels, nothing from your image layers.