skip to content

Your SSDF assessment finds every PO.5 secure-environment task sitting with a CI vendor nobody owns - what do you do?

level: principalimportance: nice to knowfreq 29%

answer

  1. The gap is an owner, not a control
  2. Operation delegates, accountability does not
  3. Every managed platform has a boundary
  4. Your side of the line holds the real gaps
  5. Never record it as not applicable

basics

~20 s

Treat it as an ownership gap, not an absence of controls. Name an internal owner for the vendor relationship and map the vendor's evidence onto the PO.5 tasks. You may outsource the operation, never the attested outcome.

solid answer

~50 s

The finding is that nobody in the company can answer for the build environment, which is worse than a missing control because it is invisible until someone asks. SSDF PO.5 asks you to implement and maintain secure environments for development; running that environment on a managed service is legitimate, but the outcome remains yours to attest. So: assign a named internal owner for the vendor relationship, obtain the vendor's own security documentation and attestations, and map them task by task onto PO.5 - marking each as inherited-with-evidence, inherited-without-evidence, or retained by you. What you configure yourself inside their platform never transfers, and that split is usually where the real gaps are. Anything left unevidenced becomes an explicit decision with a named risk owner and an expiry date, not a blank. For a payments platform the exposure is concrete: a build environment you cannot vouch for is a path to the artifacts that move money.

go deeper

for a junior

Take away one rule: using a managed service does not remove the practice from your assessment. Somebody in your organisation still has to be able to show it is handled.

for a middle

Be able to draw the responsibility boundary - what the platform provider secures versus what you configure inside it - and explain why the second half is where gaps usually hide.

for a senior

Show how you would map vendor evidence onto individual tasks, mark unevidenced ones honestly, and convert each into an in-house control, a contract commitment or a dated accepted risk.

for a principal

Own the organisational half: who funds vendor security review, what leverage renewal gives you, the concentration risk a single build provider creates, and the discipline of attesting only to what you can prove.

## What the finding actually says A gap assessment that returns 'PO.5: delegated to the CI vendor' and stops has found something more interesting than a control weakness. The controls may well be excellent - a serious managed build provider very likely runs a tighter environment than the company would run itself. The defect is that **no one inside the organisation owns the relationship**: nobody reviews what the vendor claims, nobody is told when it changes, nobody renews anything, and nobody can produce evidence when a customer or an assessor asks. An unowned dependency is a decision that was never made. The distinction to hold on to: **operation can be delegated; the attested outcome cannot.** When you sign a statement that your development environments are secured, you are answering for the whole outcome regardless of whose hardware it runs on. 'Our vendor does that' is not an answer to an attestation; it is a description of where the evidence has to come from. ## Working the finding **1. Name an owner first.** Before any technical work, someone accountable is assigned to the vendor relationship - typically the platform or build engineering lead, with procurement and security as stakeholders. This single step converts a blank into a tracked item, and it is the part organisations skip. **2. Split PO.5 into inherited and retained.** Every managed platform has a boundary. Roughly: the vendor owns the environment's isolation, patching, tenancy separation and physical and operational security; you own what you configure inside it - which identities can trigger a build, what those identities can reach, what credentials are available to a job, and who can change the build definition. Almost every real gap in this pattern lives on *your* side of that line, and it is invisible while the whole practice is written off as 'the vendor's'. Write the boundary down; it is the artifact that makes the rest of the assessment possible. **3. Collect evidence for the inherited half.** The vendor's own security documentation, independent audit reports, published attestations and contractual security commitments. Map each to the PO.5 tasks it actually covers and mark honestly where nothing covers a task. A vendor claim you have not read is not evidence. **4. Record what is left.** Every unmatched task becomes one of: a control you bring in-house, a commitment you negotiate at renewal, or an explicitly accepted risk with a named owner and a review date. What it must never become is a blank cell, because a blank is indistinguishable from an oversight, and next year nobody will remember which it was. ## The organisational layer This is where the question stops being technical. - **Nobody funds it.** Vendor security review is unglamorous work that shows up on no roadmap. If it is not somebody's named responsibility with time attached, it will not happen, and the finding will reappear verbatim next year. - **Procurement is where the leverage is.** Evidence commitments, notification-on-change and audit rights are cheap to obtain at contract signature and nearly impossible to obtain afterwards. If the relationship is already unowned, renewal is the one moment when it can be fixed. - **Concentration risk becomes visible.** Once someone owns the relationship, the question 'what happens if this vendor is unavailable, or breached, or changes its terms' finally has an addressee. For a payments platform that is a business-continuity question, not just a security one. - **Say what you can prove.** The pressure in these situations is to answer the questionnaire green because the vendor is reputable. Reputation is not evidence. A supplier who says 'this task is inherited from our build provider, here is their report, and these two sub-tasks are ours and currently unevidenced' is more credible than one claiming uniform green, and materially safer for whoever signs. ## What not to do Do not mark PO.5 **not applicable**. Not applicable means the task genuinely does not pertain to your software; a build environment you use every day always pertains. Do not assume the vendor's attestation covers your configuration - it covers their platform, and it explicitly does not cover what you set up inside it. And do not let the finding turn into a project to bring the build platform in-house, which is usually the most expensive possible answer to a documentation-and-ownership problem. The specific technical hardening of a build runner is a separate discussion with its own answers; what the SSDF assessment is asking here is narrower and harder to dodge - who in this company can show that it is done.

  • Can you mark PO.5 as not applicable because you run no build infrastructure yourself?
    No. Not applicable is for tasks that genuinely do not pertain to your software, and a development and build environment you use daily always pertains. The correct treatment is inherited-with-evidence, with the portion you configure inside the vendor's platform kept explicitly on your side of the boundary.
  • The vendor sends an audit report covering their platform. Does that close the practice?
    It closes the inherited half at best, and only for the tasks it actually addresses - read it and map it rather than filing it. It says nothing about your configuration inside the platform: who can trigger builds, what a job can reach, and who can alter the build definition. Those stay yours.
  • How do you keep this from reappearing identically at next year's assessment?
    Attach the vendor relationship to a named role with time allocated, put evidence and notification-on-change commitments into the contract at renewal, and set a review date on every accepted risk. Findings recur when they are recorded as facts about the world rather than as items with an owner and a date.

Renting a warehouse does not make the goods inside somebody else's problem. The landlord secures the building; you still answer for who has keys to your unit and what you left in it.

saying these in an interview costs you the question

  • Marks the practice not applicable because a vendor runs it
  • Treats a vendor attestation as covering your configuration
  • Says the vendor is reputable, so the outcome is met
  • Leaves the relationship without a named internal owner
  • Proposes insourcing the platform to fix a documentation gap

context