Principal engineering, through to production
Cloud, integration and software built to run. One principal engineer, 26 years.
Website · Diagnostics · Reference · Skills · LinkedIn · Contact
DBHQ Consulting Ltd provides engineering delivery services led by Daniel Grimes - 26 years building software for organisations where failure was expensive, across defence, banking, insurance, energy, commodities, healthcare and identity. Cloud platforms, systems integration, data and application delivery, taken through to production and handed over.
Need principal delivery? Engage DBHQ directly, or bring it into your consultancy's programme under your brand. Remote-first from the UK.
Everything below is public and free to use.
Remote, captured, auditable execution on a machine you cannot log into
Someone can reach the machine, you cannot, and you are the one who knows what to ask it. You publish a step, it runs on the far side, and the whole run comes back as a log with every line stamped in UTC - so a hang reads as a gap rather than as slow progress, and the log returns whether the run passed or failed. Nothing is installed on the far side: the station is plain bash the operator can read before running it, and it does not tunnel, proxy or hold a connection open to anything.
A toolkit rather than a single tool, and the agent skill is one part of it:
- CLI - one static Go binary, no runtime, carrying the station payload it was built with
- Station - the far side, in plain bash. Bash 4, git and coreutils, and no Go will ever appear in it
- Relay - both sides dial out over ordinary HTTPS, and it cannot read a byte of what it carries
- MCP server - the same commands as typed tools, for an agent running the investigation itself
- Agent skill - drives the CLI from Claude Code or Codex, and installs from the marketplace below
heliograph-io/heliograph · docs.heliograph.io · heliograph-relay · MIT
Fourteen of them, built from real delivery work and given away. Every one is documented at skills.dbhq.uk. Add the DBHQ marketplace once, then install any of them:
/plugin marketplace add dbhq-uk/marketplace
| Skill | What it does | |
|---|---|---|
 |
outlook | Microsoft 365 mail and calendar from your agent: inbox, send, drafts, attachments, availability - plus PST archives extracted to markdown |
 |
trello | Boards, lists and cards end to end, plus a board digest, a due-date radar for what is slipping, and smart sorting |
 |
legwork | Decision research where every claim states how well it is supported - and a run that settles nothing says so |
 |
dovetail | Does your repository still agree with itself? Dead links, orphaned files, docs describing behaviour the code no longer has |
 |
verve | Strips AI tells from prose and puts a human voice back, in British English. Every figure, name and claim survives untouched |
 |
vela | Compiler-exact code search over a SCIP index: definitions, references, callers, change impact. .NET natively through Roslyn, so Razor and Blazor are indexed; every other language through its own SCIP indexer |
 |
garmin | Body Battery, sleep, activities and training load, answered in the conversation - or archived to markdown you keep. Credentials never leave your machine |
 |
imager | Generate and edit images with OpenAI's GPT Image 2. Iterate on drafts at a fraction of a penny, then pay once for the one you approved |
 |
gitview | Which branches are finished and safe to delete, computed by test-merging rather than inferred from a commit count a squash-merge makes meaningless |
 |
atlassian | Jira issues raised and read, and Confluence pages searched, written and updated - over the REST API with an API token and no MCP server |
 |
pennyblack | Put a PDF in the post. Printed in the UK, delivered by Royal Mail, priced and previewed before it spends, with the tracking number handed back |
 |
buildwork | Runs a repository's open issues as parallel agents - one per issue, each in its own worktree and its own pull request - then proposes a merge order. Never merges |
 |
deskwork | Files what an agent noticed as a tracked issue, keeps the dependency graph honest, and writes a roadmap into git that you can disagree with |
 |
groupwork | Puts a second agent on the work, as an adversary or as a partner. What each pattern is refused is enforced in code, so a citation cannot overclaim |
 |
headwork | Puts the decision blocking you as one question, in plain English, with every option carrying its reasoning and one of them named as the recommendation |
They install from the marketplace in one command, as does heliograph's skill - /plugin install <name>@dbhq. Full documentation, and what each one deliberately does not do, is at skills.dbhq.uk.
Public reference projects - the DBHQ approach in code:
- fca-handbook-assistant - regulated-grade AI on Azure: grounded, cited compliance Q&A over the FCA Handbook. Full-stack .NET reference implementation
- ai-landing-zone-starter - a secure Azure AI landing zone on Microsoft's official AVM module, with a decision log and a costed deploy and teardown
- devcontainer-devops - the whole DevOps toolchain in one container: Terraform, Terragrunt, Azure CLI, Kubernetes, Ansible, PowerShell, .NET and the coding agents. Built natively for Intel and Apple Silicon, rebuilt weekly, and every image ships an SBOM and signed build provenance you can verify
- spreadsheet-health-check - is the spreadsheet your business runs on a risk? 17 checks for key-person dependency, hidden errors, macros, external links and the point where it stops scaling. Runs entirely in your browser, nothing is uploaded. Every release is hashed and provenance-attested, so you can verify the exact bundle that scored your workbook. Run it free
- azure-war-taster - a headline Well-Architected score for an Azure estate, from Azure Advisor signal and the Defender secure score. Deterministic, read-only, no sign-up, and nothing leaves your tenant. It is honest about its ceiling: 40+ of Microsoft's 59 checklist items cannot be read off a machine, and it says so in every report. The full review
