Inventory the shipped system
Start with the packages and versions that reach the deployed application, including indirect dependencies where the ecosystem exposes them. GitHub’s dependency graph can help inspect dependency relationships. Keep the inventory tied to a revision and lockfile rather than a developer’s current machine. A list produced after several unrecorded updates may not describe the version customers are using. Include build and deployment tooling in a separate view of the review.
Turn alerts into questions
For each important alert, establish the affected version, advisory conditions and whether the relevant functionality is used. Do not dismiss an alert simply because the package is indirect, or assume every alert has identical exposure. Record what is known and what needs verification. The useful output is an explanation of risk and a next action, not a promise that a zero-alert dashboard means the whole application is secure.
Examine the upgrade path
Identify whether a correction requires a small update, a breaking change or replacement of an unmaintained component. Read the relevant upstream release information before changing production. For an authentication library, verify session and permission behaviour after the update rather than limiting the test to whether the app starts. Keep the previous deployment and recovery plan available in case the change affects a critical customer journey.
Use a representative review record
For a package used in invoice generation, record why it is present, which process calls it and what inputs it receives. Link the advisory and selected correction. Then record the test demonstrating that invoice generation still produces the intended result. This small record gives a future maintainer more context than “dependency bump” and helps distinguish the deliberate correction from unrelated changes bundled into the same release.
Assign ongoing ownership
Decide who reviews new alerts and how urgent findings reach the release queue. Schedule dependency work according to exposure and maintenance needs rather than waiting for a launch deadline. Retain reproducible builds and avoid casual version changes during an incident. The long-term benefit comes from an inventory, an owner and a tested update path; the scanning tool supports that process but does not make the decision for the team.
Sources & further reading
Reference material for the guidance and examples above. Where included, community discussions provide context rather than verified incident evidence.
Retrospective weekly fieldnote, prepared with AI assistance and published on 12 September 2026. Examples are illustrative; this is not a client case study or a claim about events in the assigned week.