Setting the file. One moment.
Skill 180 · Foundry Agentic Loop
Subchapter 180.35
evals/fixtures/improve-history/AGENTS.mdMarkdown1 KBView on GitHub
This workspace is built with spec-driven development. For build requests and the short commands below, load the foundry-agentic-loop skill and follow its references/sdd.md; it checks the live run first and resumes a running or paused run instead of starting a second one.
| Command | Step |
|---|---|
status | Report the live run from dynamic_workflows_manage, then the gate issues in its .agentic-loop/runs/ record |
approve, resume | Resume the paused run with resumeFromRunId |
deploy | If a run is active, deploy only through it: at approve-deploy-v1 or approve-deploy-r<n>[-a<k>]-v1, summarize it and resume it once approved; at any other pause, report where it stopped and resolve that first. With no active run, run the Deploy phase guide, which asks before azd up |
specify, implement, verify | With no run active, run that phase guide on the existing artifacts |
build <brief> | Offer to resume a running or paused run; otherwise start a new foundry-agentic-loop-workflow run |
improve | Improve mode |
docs/spec.md, lessons in docs/LOOP.md and Azure settings only in .azure/deployment-plan.md, never here.