Engineering / Field note
A green build can still leave release security unknown
A successful build, a dependency review and a compatible production runtime answer different questions. A release decision becomes clearer when missing evidence stays visible instead of borrowing a neighboring green result.
A build result is easiest to trust when the question behind it is explicit. An application can compile, pass functional checks and produce an archive while important release-security questions remain unanswered. In our view, the difficult failure is often the missing check that inherits the confidence of a neighboring green result.
Consider a hypothetical customer portal whose ordinary build succeeds while its advisory service is unreachable. The application may be perfectly capable of running. Its current dependency-vulnerability status is still unresolved. Treating those two statements separately gives a release owner a useful decision instead of a cheerful color with an incomplete meaning.
Three Composer checks, three different questions
Composer’s validate command checks the manifest and, when a lockfile exists, whether that lockfile is up to date with the manifest. Its strict option also makes warnings affect the exit status. This is a consistency check, rather than a dependency-advisory assessment. [1]
The audit command checks dependency policies, including vulnerability advisories. Its locked option examines lockfile packages regardless of what is installed. The command reference also documents an option to ignore unreachable repositories or policy sources, making the selected policy part of the meaning of an audit result. [1]
The check-platform-reqs command compares PHP and extension requirements with the real platform and ignores simulated config.platform settings. By default it evaluates installed packages; a separate lock option changes that scope. Compatibility evidence therefore needs to identify the environment and package set being checked. [1]
Those distinctions matter when a release crosses team boundaries. A developer may be answering whether the dependency description is internally consistent, while an operator needs to know whether the target runtime can support it. A security reviewer may be asking what current advisory data says about the selected packages.
Our interpretation is that each answer should travel with its scope. A passing consistency result cannot supply a missing advisory lookup, and a successful compatibility check cannot establish that a package has no relevant security advisory. The tool names are short; the decisions they support deserve more precise labels.
Missing evidence is an operational state
There is a practical difference between a check that ran and found no relevant finding, a check that was deliberately excluded, and a check that could not obtain its input. Preserving that difference helps an owner decide whether to stop, investigate or accept a documented exception. Collapsing all three into success removes the decision.
A synthetic exercise can test whether a release process notices missing evidence. It can also test how an exception is recorded or who acknowledges a blocking result. Those are useful process outcomes. They do not establish the safety of actual dependencies, the contents of an actual deployment or the delivery of a real operational alert.
For a hypothetical release package, a functional test might exercise a payment screen while an unrelated bundling step copies internal configuration into public assets. That example illustrates why the deployed artifact has its own questions. It is not a report of a customer incident or a recommendation for a particular secret-scanning product.
OWASP’s secrets-management guidance covers storage, access, auditing and rotation, including how secrets interact with CI/CD tooling. That scope extends beyond searching source files for suspicious strings. The practical concern is how a secret is provisioned, who can use it and where it can appear during a release. [3]
The runtime has its own maintenance clock
As checked on November 18, 2026, PHP 8.2 remains in its security-support period, which ends on December 31, 2026. The PHP project publishes distinct active-support and security-support dates. That is a planning deadline alongside the separate task of identifying the patch level actually running. [2]
In our view, a release decision benefits from keeping the selected source revision, dependency set, deployed artifact and target runtime connected. An audit result from a different revision may be useful history, but it leaves a question about the release currently being authorised. A record without that connection asks the next person to reconstruct it.
The same applies to exceptions. An accepted risk has an owner, an understood scope and a condition for revisiting it. An unavailable check has an unresolved input. Recording them accurately can keep a release moving where appropriate without making the evidence say more than it does.
For the people receiving the application, the useful outcome is a clear account of which security questions were answered for this release and which remain open. A green build can remain a good build result. Giving dependency, runtime and secret-handling evidence their own meaning makes the overall release decision more honest.
Sources + further reading
Source dates describe the original material. Publication above is the date this Zyel article went live.
- Composer command reference: validation, dependency audit and platform requirementsComposer · Checked Nov 18, 2026
- PHP supported-version maintenance periodsPHP project · Checked Nov 18, 2026
- Secrets Management Cheat SheetOWASP · Checked Nov 18, 2026
Reader discussion
What would you ask next?
Questions, useful corrections and thoughtful disagreement are welcome. Comments are reviewed before they appear.
No reader comments yet.