[kumo-isaac]

Summary

@devanj08, this is a 4-line follow-up to #154, stacked on docs/graph-visualization-skills. It fixes two codegen problems the evals found in #154. The router and intake are untouched.

What happens and why

The codegen eval (kumo_skill_dev harness, 38 cases scored against reference specs) showed two side effects of the new graph-preview guidance:

1. The agent added a primary key to an event table (primary_key="STATUS_EVENT_ID" on rig_status_events) in 3 of 4 runs on Surface graph visuals in customer agent workflows #154, and in 0 of 4 on its base. The viewer doesn't need it: render() only requires keys on edge destination tables, which Graph already enforces.
2. Notebooks also saved graph.svg (16 of 38), although the SVG export is meant for scripts only.

The fix

Four lines in kumo-graph-visualization/SKILL.md:

• The visual shows the graph exactly as built: no primary key, column or edge is added for the viewer, and keyless tables display as they are.
• A notebook displays the graph only; it saves no SVG unless the user asks for a file.
• The script section applies to scripts only.

Validation

Base fa46bed, #154 head 86eccb1, and this PR 52b00016, all loaded through the same eval loader. The loader was patched to include kumo-graph-visualization/SKILL.md in the codegen prompt, because the router tells every train notebook to use it.

	base	#154	this PR
Codegen spec match (38 cases)	38/38	36/38	38/38
PK case, no key on event table (4 runs)	4/4	1/4	4/4
end_date case (4 runs)	4/4	3/4	4/4
Notebooks with the preview cell	0	38	38
Notebooks saving graph.svg	0	16	0
Intake prompt sha	34e889459f9c	94f094938e37	94f094938e37 (identical, so intake unchanged)

Intake on #154 itself: 82/90 against 80/90 on base (Nemotron judge), with no case regressed and no training request routed to the visualization skill by mistake.

Why the router is not touched: rewording its SVG sentence to "A script, never a notebook, includes…" fixed codegen equally well, but it broke two intake cases where a field is missing. nov- and videoamp-missing-positive-interaction asked back 0 of 3 times, against 15 of 16 on base and #154. Putting the change only in the visualization skill, which intake never reads, avoids that.

agent-skills tests: 184 passed.

Not covered:

• The eval has no script-output cases, so the script SVG path isn't eval-tested.
• No GPU run.

Blast radius

Skill guidance only, in one file that the agent reads at codegen time.

Separately, for #154 as a whole: no released kumo-engine build has Graph.visualize. PyPI 1.0.0 (2026-10-17) predates #134. Until a release includes it, the preview cell prints "Graph preview unavailable". Details: #154 (comment)

🤖 Generated with Claude Code

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on base/target branches other than the default branch.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

Run configuration

• Source mode: Auto
• Repository: NVIDIA/kumo-engine
• Branch: docs/graph-visualization-skills
• File: .coderabbit.yaml
• Commit: 65a0711
• Layers: Repository YAML
• Review profile: QUIET
• Plan: Enterprise
• Run ID: 7bd04f0c-9d18-460a-97a7-e3fe3bfb137d

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• Trigger review

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts