Quoted by complexity
CI/CD & quality-gate setup
One deterministic rule set - format, lint, types, tests, coverage, security - that runs identically in the editor, before push and in CI. Ratcheted so legacy code can adopt it on day one.
For AI agents: a documentation index is available at /llms.txt. Every page also has a Markdown version at its URL plus .md.
← What I do
~/what-i-do/devops
DevOps & CI/CD
A pipeline is a promise about quality that runs on every change. I build ones that are deterministic, fast, cheap to run and hard to quietly weaken, including by AI agents.
01What I can do for you
Hourly or quoted
Legacy modernisation
Framework and stack upgrades, deprecated code removed and duplicated logic consolidated, with tests and a quality gate in place first, so the upgrade doesn’t break what already works.
02How I work
One rule set, three placements
In-loop, pre-push and CI enforce the same rules. Different rules in different places is how you get "green locally, red in CI".
Seen in: Foundry →
The exit code decides
Linters, types, tests and scanners own pass/fail. AI may propose a fix, but it only lands when the deterministic gate passes.
Seen in: Foundry →
Secure the supply chain by default
SHA-pinned actions, least-privilege tokens, secret and dependency scanning, and no long-lived keys in CI.
Seen in: Foundry → · jev-tools →
Nobody quietly weakens the gate
Loosening a rule is its own pull request with a label a human sets. A change that relaxes the gate next to the code that needed it is refused, whether a person or an agent wrote it.
Seen in: Foundry →
Red CI is a task, not a report
Every push is watched until it is green. A failure gets its cause fixed and pushed again, not a retry button or a note saying it’s probably unrelated.
Small, single-purpose changes
Every commit builds and does one thing. Refactors and formatting ship apart from behaviour changes, so a review reads the change, not the noise.
Measure the pipeline
Know where the gate spends its time and whether a change is converging or thrashing, so the slow step gets fixed instead of tolerated.
03What I’ve built
04Writing
A quality gate for a codebase nobody dares touch - a one-week plan
How to add linting, tests and security scanning to a legacy repository without a cleanup sprint, a red build on day one, or a revolt.
Case study: a cheap, typed model in front of the expensive one
Three open-source tools - an agent guardrail, a GitHub issue-triage Action and a bulk classifier - built on one pattern for AI automation that's cheap enough to run on everything.
One rule set, three placements
Why I stopped treating CI and "AI coding standards" as two separate things - and what Foundry does instead.