What problem does it solve?
Grant pull requests in the marin repository use opaque human-NNN references whose emails are KMS ciphertext in iam_data.yaml, so reviewers cannot see who actually receives access. This Skill decrypts the changed principals, enforces an explicit human confirmation of the real identities and roles, then merges and applies the grant with pulumi up.
Core Features & Use Cases
- Decrypt grant diffs: Runs
iam_principal.py decrypt --diff against the PR diff to reveal the real user:<email> principals behind opaque references.
- Human-in-the-loop confirmation: Presents a plain-language summary of each added or revoked grant and refuses to approve or merge without explicit user approval.
- Safe apply workflow: Approves and squash-merges the PR, then runs
pulumi preview and pulumi up on the affected stack, stopping on any unexpected drift such as NodePool replacements.
- Use Case: A teammate opens PR #1234 adding a storage viewer grant. Use this Skill to check out the PR, decrypt that it grants [email protected]
roles/storage.objectViewer, confirm with the requester, merge, apply the marin stack, and comment on the PR that the grant is live.
Quick Start
Review grant PR 1234 in marin-community/marin, decrypt its principals, and after my confirmation merge it and apply the marin Pulumi stack.