Your product includes code you did not write
Modern apps are mostly dependencies. That is normal — and it means attacker effort often targets the package ecosystem and build pipeline, not only your application code.
Buyers now ask what you ship, how you verify it, and how fast you can remove a bad package. “We use npm audit sometimes” is no longer a complete answer.
Supply chain hygiene is delivery work: lockfiles, reproducible builds, signed artifacts where it matters, and a patch habit that survives busy sprints.
Ignore it and you will discover the gap during a customer questionnaire or a public CVE week — both expensive times to start.
Know what you ship: SBOMs and lockfiles
Generate a software bill of materials from the build that actually deploys, not from a developer laptop. Store it with the release.
Commit lockfiles and fail CI when they drift. “Latest” floating ranges are convenience for attackers and chaos for incident response.
Track base images and language runtimes the same way. An old distroless or Node image is still supply chain surface.
When a customer asks “are we affected?”, you need a queryable inventory — not a Slack archaeology project.
Protect the build, not only the repo
Compromised CI credentials and overly powerful deploy keys bypass application review. Treat pipelines as production infrastructure.
Pin GitHub Actions or equivalent by digest where practical. Review new marketplace actions like you review new dependencies.
Separate who can change pipeline definitions from who can merge feature code when your risk justifies it.
Secrets in CI should be short-lived and scoped. Permanent cloud keys in pipeline variables are a classic blast-radius mistake.
Policy that does not freeze delivery
Block known-critical issues on the path to production, with a documented exception process that has an expiry date.
Batch routine upgrades on a cadence so you are not always firefighting. AI-generated code that adds new packages needs the same review as human PRs.
Prefer allowlists for high-risk ecosystems or internal mirrors when customer contracts demand it — but start with visibility and fast patching if you are early-stage.
Measure mean time to remediate critical dependency CVEs. That metric matters more than a perfect score that nobody maintains.
Respond when a package goes bad
Have a runbook: who watches advisories, how you find affected services, how you ship a hotfix, and how you tell customers.
Practice once with a fake CVE. Note where inventories are incomplete or deploys are too slow.
Keep rollback paths boring. Supply chain incidents often need a previous known-good artifact, not a heroic rewrite.
After the event, fix the gap that made detection slow — do not only patch the one library.
What to put in security questionnaires and briefs
Describe how you generate SBOMs, how CI enforces dependency policy, and who owns patch SLAs. Link to real process, not aspirational policy.
When briefing a partner, share your package managers, container strategy and any customer contractual requirements.
Ask them to leave transferable automation in your repos — scanners and policies you keep after the engagement.
Avoid one-off manual audits with no CI follow-through; the next dependency lands tomorrow.
How Three Index hardens the path to production
Our DevOps and cloud work includes pipeline hygiene, artifact traceability and operational habits that survive real release pressure.
We favour controls your team can run weekly over a glossy report that ages out in a month.
Combined with delivery teams in your repository, upgrades and fixes stay part of normal shipping — not a side project.
Send a brief with your stack and the supply chain questions customers already ask. We will prioritise inventory, gates and response readiness.

