Not in project documentation
Passwords should not be stored in ordinary project documentation, tickets, chat threads or shared notes, where they outlive their purpose and spread beyond the people who needed them.
KRUQ Trust Center 2026
A practical overview of how KRUQ approaches security, privacy, access management, responsible AI and operational control throughout technology projects.
Access is limited to what a project requires.
Only information necessary for delivery should be requested.
Changes are versioned, tested and released deliberately.
AI-assisted work remains subject to human review.
Access and change are the two places where most risk enters a project. Both are handled deliberately, from the first credential request through to production release.
Permissions are requested against a defined task rather than as broad administrative access. Where a narrower role achieves the same outcome, the narrower role is used.
Individual named accounts are preferred over shared logins so that activity remains attributable within the client's own systems and audit trails.
Access granted for a specific phase is intended to be withdrawn once that phase ends, and remaining access is reviewed as the engagement closes.
Work is tracked in version control so that changes have history, authorship and a reviewable record rather than arriving as untracked edits.
Changes are exercised in a non-production context appropriate to the project before they are proposed for a production environment.
Releases are planned, scheduled and reversible where the platform allows, rather than applied ad hoc to live systems.
Credentials are treated as sensitive material with their own handling rules, separate from ordinary project documentation.
Passwords should not be stored in ordinary project documentation, tickets, chat threads or shared notes, where they outlive their purpose and spread beyond the people who needed them.
Secrets should not be committed to source-code repositories. Repository history is long-lived and widely readable, so a committed secret should be treated as disclosed even after removal.
Where the platform or client environment offers secret management — a vault, a managed secret store or environment-level configuration — those methods should be used in preference to manual handling.
If exposure is suspected, the credential should be rotated rather than assessed for likelihood of misuse. Rotation is treated as the ordinary, low-drama response.
The safest data is the data a project never needed to hold. Handling starts by narrowing what is requested in the first place.
Data accessed during an engagement is used for the delivery purpose it was provided for, and not repurposed for unrelated activity.
Access is scoped to the systems and records a task genuinely requires, rather than to whatever the client's role model makes convenient to grant.
Where a task can be completed with anonymised, masked, sampled or synthetic data, that is preferred to working against a full production dataset.
Where appropriate, production environments and the data inside them remain under the client's own control and administration, with KRUQ working through the access the client grants.
Project closure includes reviewing what access, accounts and copies of data still exist, so that nothing quietly persists past the end of the work.
Modern software is assembled as much as it is written. What a solution depends on is part of the solution.
APIs, cloud services, libraries and external platforms are treated as explicit project dependencies rather than invisible implementation detail. Their role, and the implications of relying on them, should be considered during solution design — while the choice is still open — instead of being discovered after a system is already built around them.
AI is used as an engineering tool inside a controlled process — not as an unreviewed author of client deliverables.
AI-assisted work is reviewed before becoming a client deliverable or production change. A person remains accountable for what is handed over.
Confidential client information should not be submitted to AI systems without appropriate authorisation and safeguards. Where in doubt, the default is not to submit it.
AI should be used where it creates genuine project value and should not blindly replace human judgement. The measure is the quality of the outcome, not the novelty of the method.
Each stage has its own entry point, its own controls and its own exit. Access and change are granted stage by stage, not all at once.
Scope, constraints, environments and data sensitivity are established, along with what access the work will actually require.
Least-privilege access is requested against that scope, using named accounts where supported and the client's own approval process.
Work proceeds under version control with review, in non-production environments, with secrets kept out of the repository.
Changes are tested, then deployed deliberately — planned, agreed with the client and reversible where the platform allows.
Documentation, ownership and credentials transfer to the client, and access that is no longer required is reviewed and removed.
Governance is what keeps a project legible: what was agreed, what changed, and how anyone can tell whether it is on track.
Scope, deliverables, assumptions and responsibilities are written down before delivery begins, so both sides are working from the same document.
Work is delivered against the agreed scope. Items outside it are identified as such rather than absorbed silently into the schedule.
Changes to scope, approach or timeline are raised explicitly, with their impact on effort and delivery made visible before they are accepted.
Deliverables are defined so that "done" is verifiable — a shared, stated standard rather than a matter of impression.
Progress, blockers and risks are reported during the engagement, while there is still time to act on them.
Closure is a defined step: documentation, ownership, credentials and access are transferred and reviewed rather than left to lapse.
Position
KRUQ believes trust should be based on clear practices and verifiable commitments rather than unsupported marketing claims.
Security & Trust
For vendor onboarding, due diligence or project-specific security requirements, contact KRUQ. Requirements can be reviewed during discovery and documented in the applicable proposal, Statement of Work or security annex.