flowlever

Write a spec workspace's ACCEPTED change proposals back to their real sources — Azure DevOps work-item fields and Confluence pages — surgically and only on the user's explicit Apply. The spec-side counterpart of posting PR comments: it reads each finding's reviewed draft from the ledger and applies the accepted/edited ones, patching a single field/section without rewriting the rest of the page. Use when an `apply` request targets a `spec` workspace (the /flowlever:watch runner dispatches it), or when the user says "apply the accepted spec changes for X".

Run a spec-readiness audit for a feature across its Confluence spec, Azure DevOps work items, and Figma designs, then ingest the findings into the FlowLever ledger. Use when the user says "audit feature X", "/flowlever:audit", "validate the spec for X", "is X ready for implementation", "check spec readiness", or shares a Confluence/ADO/Figma link and asks whether it's consistent / complete / ready to build.

Compose an implementation-ready brief for a feature once its FlowLever readiness gate is green, pulling together the spec, work items, designs, resolved decisions, and known tradeoffs. Use when the user says "implementation brief for X", "hand off X to the devs", "is X ready to build", "/flowlever:brief", or "write the build brief for X".

The scheduled autopilot pass for PR reviews. Discovers Azure DevOps PRs where you are a reviewer (plus reviewer threads on your own PRs), decides per PR whether to review, re-review, or skip — aware of reviews you triggered manually — enqueues the work on the cockpit queue, then drains it like /flowlever:watch. Ingest-only: it NEVER posts to a PR (posting stays behind your explicit Apply). Use when the user says "/flowlever:poll", "run the review poller", "check my PRs for reviews", or from a scheduler (cron/launchd running `claude -p "/flowlever:poll"`).

Load the reviewer feedback on YOUR Azure DevOps pull request into the FlowLever cockpit as a `pr-respond` workspace: pull the comment threads that await your response, ingest each as a finding with a proposed reply/fix, so you can step through and decide thread-by-thread in the same review UI — then post the replies (and apply code fixes) back. Use when the user says "respond to PR X in the cockpit", "load my PR threads into flowlever", "/flowlever:pr-respond X", or "work through PR X feedback". For a plain interactive/one-shot response without the cockpit, use the standalone `/pr-respond` skill.

Load an Azure DevOps pull request into the FlowLever cockpit as a `pr-review` workspace: fetch the PR diff + spec/ticket, produce review findings anchored to file:line, and ingest them so you can step through and decide each one in the same review UI as specs — then post the kept comments back to the PR. Use when the user says "review PR X in the cockpit", "load PR X into flowlever", "/flowlever:pr-review X", or "PR review X with the stepper". For a plain markdown/interactive review without the cockpit, use the standalone `/pr-review` skill instead.

Draft concrete change proposals for a FlowLever spec workspace's open findings — before→after edits to Azure DevOps work-item fields and Confluence spec sections — and attach them to the ledger so the user can Accept / Edit / Reject-with-counter each one in the cockpit, exactly like PR-review comments. Read-only: it never writes to ADO/Confluence (that is /flowlever:apply-spec, on the user's explicit Apply). Use when the user says "propose fixes for X", "draft the spec changes", "/flowlever:propose X", or after an audit when they want reviewable change drafts rather than just a findings list.

Work through a feature's open FlowLever findings and draft concrete fixes to the Confluence spec and Azure DevOps work items, applying them only with the user's explicit confirmation. Use when the user says "rework findings for X", "fix the spec issues", "address the open findings", "/flowlever:rework", or "let's clear the blockers on X".

Launch the FlowLever cockpit — start the local dashboard server and open it in the browser — so the user can work entirely from the UI. Use when the user says "start lever", "/flowlever:start", "open the dashboard", "launch the cockpit", "run flowlever", or "open the spec dashboard".

Stop the FlowLever cockpit — kill the local dashboard server and cleanly end the /flowlever:watch loop. Use when the user says "/flowlever:stop", "stop flowlever", "stop the cockpit", "kill the server", "shut down flowlever", or "stop the watch loop".

Fill in what a FlowLever workspace is missing from its ALREADY-REGISTERED sources: the plain-language "what this is about" summary, and — while the work items are open anyway — the work-item types and Vertec booking phase the cockpit needs to tell a PR from its user story. Read-only toward Azure DevOps and Confluence; writes nothing to the PR and ingests no findings. Use when the cockpit's "Generate summary" button enqueues a `summarize` request (the /flowlever:watch runner dispatches it), or when the user says "write the summary for X", "summarise this PR in flowlever", or "/flowlever:summarize X".

Give a fast status read on a feature's spec readiness from the FlowLever ledger — score, what's open, what moved since the last audit round, stale and pinned findings — without fetching or changing anything. Use when the user says "what's the status of X", "where are we on X", "what's still open on X", "/flowlever:track", or "spec status".

The session-side runner for FlowLever's UI-triggered jobs. Polls the cockpit's request queue and executes each queued request — starting a PR review/respond from a PR id, or posting reviewed decisions back — by delegating to the matching /flowlever:* adapter. Use when the user says "/flowlever:watch", "watch the cockpit", "run the flowlever runner", or sets it on a loop ("/loop /flowlever:watch"). This is what lets the user kick off PR reviews from the web UI.