[All Platforms][Docs] configure-model-limits renders five empty sections across the three agent variants, and the Deep Agents page documents no way to set any limit

## Description

The "Configure Model Limits" page is published in three agent variants. Content is filtered per agent, but the section headings are not, so headings for the other agents remain on each page with nothing beneath them. Five sections render empty across the three variants.

The Deep Agents variant is the serious case. Three of its five sections are empty, and the page never names a single variable a Deep Agents user can set. A page titled "Configure Model Limits" therefore contains no instructions for configuring model limits. It even states that such values exist without ever saying what they are:

"Use this recreation path for Deep Agents context-window or maximum-token changes so the
managed runtime configuration and OpenShell route stay aligned."

The reader is told the values exist and that changing them requires recreating the sandbox, but not what they are called or where to set them.

Empty sections by variant:

openclaw : "Set the Hermes Context Window"
hermes : "Set OpenClaw Limits"
deepagents : "Set OpenClaw Limits", "Set the Hermes Context Window", "Use Detected Local Limits"

The first two rows are agent-specific sections whose bodies are correctly filtered out; only the leftover heading is wrong there. The third row is different: "Use Detected Local Limits" carries identical, agent-neutral text on both the openclaw and hermes pages, describing how a local server reports its context length, including that local vLLM and OpenAI-compatible endpoints can report max_model_len through /v1/models. That guidance applies to Deep Agents as well, and the Deep Agents page loses it entirely.

The content that Deep Agents readers cannot reach exists on the sibling pages:
OpenClaw: NEMOCLAW_CONTEXT_WINDOW (default 131072), NEMOCLAW_MAX_TOKENS (default 4096)
Hermes: NEMOCLAW_CONTEXT_WINDOW (at least 64000, unset means auto-detect)

Two separate corrections are needed. Removing the empty headings alone would leave the Deep Agents page tidy but still unable to teach the task.

1. Suppress headings whose bodies are filtered out, for all five occurrences.
2. Supply the missing Deep Agents content: how to set its context window and maximum output tokens, and restore the agent-neutral detected-limits guidance.

Platform scope: Not platform-specific. The defect is in published page content and was confirmed in a browser against the live documentation site on the openclaw, hermes, and deepagents variants.
Regression: Unknown - the page was not validated against an earlier build.
OpenShell issue: No

## Environment

```text
Device: Not applicable - published documentation defect
OS: Not applicable
Architecture: Not applicable
Node.js: Not applicable
npm: Not applicable
Docker: Not applicable
OpenShell CLI: Not applicable
NemoClaw: Documentation for v0.0.111
OpenClaw: Not applicable
```

## Steps to Reproduce

```text
1. Open the Deep Agents variant of the "Configure Model Limits" page under the NemoClaw user guide, in the inference / manage-inference section.

2. Read the whole page and try to determine which variable sets the Deep Agents context window or maximum output tokens.

3. Observe that the sections "Set OpenClaw Limits", "Set the Hermes Context Window", and "Use Detected Local Limits" have headings but no content, and that no variable name appears anywhere on the page.

4. Open the openclaw variant of the same page. Observe that "Set the Hermes Context Window" is present but empty.

5. Open the hermes variant. Observe that "Set OpenClaw Limits" is present but empty.

6. Compare "Use Detected Local Limits" on the openclaw and hermes pages. Observe that the text is identical and agent-neutral, and that it is absent from the Deep Agents page.
```

## Expected Result

Each variant shows only the sections that apply to it, with no empty headings. The Deep Agents page states which variables control its context window and maximum output tokens and how to set them, and retains the agent-neutral guidance about limits detected from a local server.

## Actual Result

Five sections render as headings with no content across the three variants. The Deep Agents page contains no variable name and no procedure, so it does not document the task named in its own title, while still referring to Deep Agents context-window and maximum-token changes as though they had been explained.

## Logs

```text
# Deep Agents variant, complete body as published
Configure Model Limits

Configure explicit model limits before onboarding so NemoClaw can bake them into the sandbox image.
Changing an explicit build-time model limit on an existing sandbox requires fresh recreation.

Set OpenClaw Limits
Set the Hermes Context Window
Use Detected Local Limits
Recreate an Existing Sandbox

Model limits are build-time settings.
Recreate the named sandbox after changing a supported value.

nemo-deepagents onboard --fresh --name {sandbox-name} --recreate-sandbox

Use this recreation path for Deep Agents context-window or maximum-token changes so the managed
runtime configuration and OpenShell route stay aligned.

Related Topics
Configure Inference Timeouts
Switch Models

# hermes variant - same heading, content present
Set the Hermes Context Window
Hermes accepts NEMOCLAW_CONTEXT_WINDOW as its model-limit override.
NEMOCLAW_CONTEXT_WINDOW Positive integer, at least 64000 tokens Unset so Hermes auto-detects

# openclaw variant - same heading, content present
Set OpenClaw Limits
OpenClaw accepts an explicit context window and maximum output-token count.
NEMOCLAW_CONTEXT_WINDOW Positive integer in tokens 131072
NEMOCLAW_MAX_TOKENS Positive integer in tokens 4096

# Use Detected Local Limits - identical on openclaw and hermes, absent on deepagents
When NEMOCLAW_CONTEXT_WINDOW is unset, NemoClaw can use a context length reported by the
selected local server. Local Ollama reports the loaded model's runtime context length.
Local vLLM and OpenAI-compatible endpoints can report max_model_len through /v1/models.
Set NEMOCLAW_CONTEXT_WINDOW when you need to override the detected value.
```

Start at the NemoClaw user guide's inference / manage-inference "Configure Model Limits" page and compare its openclaw, hermes, and deepagents variants. Trace the variant filtering that leaves headings behind, then verify the published pages show only applicable sections, Deep Agents limit variables and procedures, and the detected-limits guidance. No source file or test is named in the issue.

Written by the indexing model from the issue text.