[danielle-mustillo]

---

Written by 🤖 (start)

Why

New installs have no example agents. Skills need kubectl today.

What changed

• Built-in Agent. platform-agent runs on kagent-default.
  • It sends Kubernetes, Helm, Istio and other questions to specialist AgentTemplates.
  • builtins.platformAgent.enabled: false turns it off.
  • builtins.enabled is off in the chart; kagent install turns it on with harnesses.enabled.
  • The chart fails when platform-agent is on without kagent-default or kagent-tools.
  • --profile minimal now skips it.
• Skills from kagent-tools 0.3.1. No separate skills image.
  • files/builtin-skills.json lists them, pinned by digest.
  • make builtin-skills-catalog regenerates it.
  • Prometheus appears only with a Prometheus URL.
• Skills in the agent template form.
  • Pick a built-in skill, or add OCI, Git or S3.
  • Each skill shows as a labeled card.
  • Old built-in versions show an Update button.
  • env-config.js now escapes multi-line values.

Limits

• Git and S3 skills need fix(substrate): let the golden actor fetch git and S3 skill sources #3435.
• Skills name tools; Agents must still bind them.
• --reuse-values upgrades render no built-ins.
• platform-agent loaded its skill in 2 of 4 chats.

Testing

1. Run kagent install on Kind.
2. Check platform-agent is Ready.
3. Ask platform-agent about a failing pod.
4. It hands off, then loads k8s-troubleshooting.
5. Open the agent template form; pick a skill.
6. Run kagent install --profile minimal; expect no built-in Agent.

---

Written by 🤖 (end)

[Charlesthebird]

This automated review is a first pass. A manual review will follow once the blockers are addressed.

---

Nothing blocks merge. One non-blocking gap: the skill form accepts an OCI image the controller rejects.

✅ What I checked

• ✅ --reuse-values from a release without builtins: every read goes through dig on .Values.AsMap
• ✅ global.imageRegistry: the catalog and k8s-agent both use the mirrored image, so "Older version" compares like with like
• ✅ Git, S3, path, 50-skill limit and duplicate-name checks against the v1alpha3 CRD and the translator
• ✅ plugins and promptTemplate on save: specFromDraft replaces only spec.skills
• ✅ KAGENT_UI_BUILTIN_SKILLS escaping through the entrypoint into JSON.parse
• ✅ CI: only the Harness named kagent clashes with the lifecycle suite, and k8s-agent drops out with kagent-tools off
• ✅ Default s3://ate-snapshots/kagent/ matches the RustFS store in setup-cluster.sh
• ✅ No internal brand terms in the diff

---

Written by 🤖

[Charlesthebird]

On second review, this LGTM overall. Couple notes:

• Could /improve-ui for both of these areas. Specifically, it could be clearer what skill the delete button is for, and the link could look more like a link.

[younsl]

The kagent-tools bump to 0.3.1 in this PR also fixes a downgrade for users upgrading from 0.10.x. The v0.10.6 chart depends on kagent-tools 0.3.1, while v1.0.0 pins 0.3.0, so moving from 0.10.6 to 1.0.0 rolls kagent-tools back.

Keeping this dependency on the latest tools release matters because operational fixes land there. For example, kagent-dev/tools#84 (the Streamable HTTP server leaking heap on every MCP initialize request) was fixed in 0.3.0 by kagent-dev/tools#85. Users only get fixes like that without overriding the subchart when the chart tracks the current release.

Changes between kagent-tools 0.3.0 and 0.3.1:

• CVE fixes in the bundled CLIs and Go dependencies (kagent-dev/tools#99)
• The kagent 1.x skills bundle in the tools image (kagent-dev/tools#91)

If the rest of this PR needs more review time, landing the dependency bump on its own would keep the next 1.0.x release from shipping the older tools chart.

[danielle-mustillo]

Thanks! Changes from your review:

1. Each skill is now a card with labeled fields and a red remove button, on the form and the details page (28f847f).
2. Went with option (c): platform-agent hands each task to a specialist AgentTemplate (Kubernetes, Helm, Istio, Argo Rollouts, Kubescape, Prometheus), each with one skill. k8s-agent stays a plain Agent (13cc6d7).

---

🤖 written by Claude

[Charlesthebird]

Thanks! I'm kicking off another automated review and video to see the changes

[danielle-mustillo]

Before is main at 7fb0935; after is this PR's images. Both are real kagent install runs on kind with an OpenAI model.

1. Agents list: a default install now creates k8s-agent and platform-agent, both Ready. On main the list is empty.
2. Templates tab: six AgentTemplates now ship. The main run has a copy of k8s-agent applied with kubectl, so the template pages below compare the same data.
3. Template details: the Skills section (AgentTemplateForm.tsx#L603) shows kagent · k8s-troubleshooting. On main only a note says the template has skills.
4. Template edit: each skill is a card with a red remove button. "Add a kagent skill" (SkillsField.tsx#L305) lists the 5 built-in skills, and "Add your own skill" adds an OCI, Git or S3 card.
5. New agent, inline template: the same skills editor now shows here too.
6. Chat: platform-agent hands off with transfer_to_agent, runs load_skill k8s-troubleshooting and k8s_get_resources, then lists the kagent pods. load_skill ran in 2 of 4 tries. The other tries listed the pods without it.
7. CLI: after kagent install, kubectl get agents,agenttemplates lists 2 Agents and 6 AgentTemplates. With --profile minimal (minimal.yaml#L4) it lists none.

pr3554-verify-reel.mp4

Screenshots

Before: Agents list after install

After: built-in k8s-agent and platform-agent

After: 6 built-in AgentTemplates

Before: template details, skills hidden

After: template details, Skills section

Before: template edit, no skills editor

After: template edit, skill card and add buttons

After: Add a kagent skill catalog

Before: New agent inline template

After: New agent inline template with skills

After: platform-agent chat answer

---

🤖 written by Claude

[Charlesthebird]

This automated review is a first pass. A manual review will follow once the blockers are addressed.

---

Nothing blocks merge. Two Medium findings are inline: the specialists bind no Kubernetes read tools, and a multi-line KAGENT_UI_BUILTIN_SKILLS override breaks the whole UI config.

✅ What I checked

• ✅ Rendered AgentTemplate and Agent fields match the v1alpha3 CRD types (tools[].mcp.server, requireApproval, subAgent.templateRef, skills[].source.oci/path, harnessRef).
• ✅ platform-agent calls k8s-agent, which has no subagents, so the compiler's one-level and no-repeat rules hold.
• ✅ helm unittest on the chart with kagent-tools 0.3.1: 35 suites, 383 tests pass; go test ./ui/entrypoint/... passes.
• ✅ UI validation matches the CRD patterns (OCI digest, Git URL and commit, S3 fields, relative paths, 50-skill cap, unique names).
• ✅ Save keeps plugins and promptTemplate; mustToJson escapes <, so a skill description cannot close the <script> tag.

Written by 🤖

[Charlesthebird]

Thanks for making these updates!

This LGTM now but just wondering if we really need 2 agents - I think just one "platform" agent should do it. And the description for the platform agent should be updated now to include that it sends "Kubernetes" questions to a specialist.

[iplay88keys]

I had thought we would have this disabled by default and only enable it by default with the kagent install flow.

Can you remind me why it's enabled by default here?

[danielle-mustillo]

They rendered nothing on a plain install because they also depended on harnesses.enabled, but that was hard to see. builtins.enabled is now false in the chart and kagent install sets it to true, like harnesses.enabled. The chart now fails with a clear message when platform-agent is on but the kagent-default Harness or kagent-tools is off.

---

Written by 🤖