Menu
SkillsMP has collected 16 skills from umbraco/umbraco-ai-ops. Open a skill to review its source and details.
Showing 16 of 16 collected skills.
Onboard a repo to the engine, and prove it is covered. Detects the repo's setup, writes the facts detection cannot reach to `.claude/ops-repo-meta.json`, reports capability coverage against the catalog (present / inherited / missing), scaffolds a stub for…
Ambient facts about this repo and its family, returned as structured data: `identity` (repo, default branch, the trigger labels), `topology` (which repo fills the `code` / `issues` / `releases` / `learnings` role), and `lines` (the live version lines, which…
Land a change that has already shipped on one line onto this repo's other live lines. Fired by the port label on a merged PR, or by `ops-merge-loop` straight after it lands one. Works out the target lines from the repo's declared port order — never from a…
Stand up the standardised loop automation for a repo — one loop-dispatch routine per repo, fired by a committed GitHub Action (not UI event triggers), with identical env, model, tools, connections, and a thin prompt every time. The Action routes at the edge…
Shared how-to for GitHub operations in both environments this system runs in — locally with the `gh` CLI + `git`, and on Claude web / in scheduled routines with the GitHub MCP server (`mcp__github__*`, no `gh` there). One operation catalog, two reference…
Event-triggered release with no mid-flow human approval, guarded by two automated gates: green CI, then an Opus pre-publish review against a repo-overridable checklist. When an issue titled `release <version>` is labelled `ops/auto-release`, it commands…
Turns a repo's `ops/ready-for-ai` backlog into CI-green PRs by commanding `ops-change` (implement → verify) per issue. It owns the loop only — the queue, the `/goal`, rolling cap-3 dispatch, model choice per issue, the CI-green cap and the backstops — and…
This repo's branch model, and the only holder of it. Merges a PR with whichever strategy the model calls for, opens a PR onto the correct base for its line, and starts a work branch rooted correctly. The integration branches, the release base and the merge…
Land an approved PR. Owns every piece of merge policy — the gates (landing label, CI green, mergeable, and a base the branch model accepts), the release-base skip, and the merge itself — and returns a structured outcome the caller reports on. A caller hands…
Sweeps the PRs a maintainer has labelled for landing and hands each to `ops-integrate · land`, then reports what came back. It owns scheduling only — the sweep, the CI poll cadence and its 15-minute cap, the cap of 10 PRs per run, and the comment on each PR.…
Single front door for a repo's automation loops, so one routine per repo can handle every loop event instead of one routine per event. Deterministically routes each triggering GitHub event (parsed from the trigger block, decided by route-event.sh) to the…
This repo's CI, read-only. `status` reports whether a PR is pending, green or red and which checks say so; `log` returns the failing part of a failing build's log, trimmed to what is needed to diagnose it. The CI provider — GitHub checks, Azure Pipelines,…
Tell a human something happened, through whichever channel this repo uses — a Claude push notification by default, or whatever a repo overrides this with. Infrastructure, not a gate: a notification that fails must never fail the work it was reporting on.…
An isolated place to build and test one change — worktree, database, ports, dependencies — and the teardown that removes everything it created. Prepared and torn down by `ops-change`, never by a loop, so how isolation is provisioned stays a product fact. The…
Label-triggered loop that acts on PR review feedback. When a reviewer has left comments and labels a PR `ops/auto-rework`, it reads every actionable point, commands `ops-change · implement` scoped to that feedback and `ops-change · verify`, pushes, replies on…
The weekly sweep that turns captured proto-learnings into reviewed work. Reads the open `ops/proto-learning` issues on the learnings repo, clusters them by lesson, applies a promotion threshold, and routes each cluster to its home — an issue on the repo that…