docs: refresh 0.0.57 release docs

[miyoungc]

Summary

• Add the missing v0.0.57 release-notes section with links to the detailed docs pages for command, inference, onboarding, messaging, status, installer, and policy changes.
• Remove public references to docs-skip terms from source docs and regenerate the NemoClaw user skills from the current Fern MDX docs.
• Carry forward generated references for the per-agent documentation split, including Hermes-specific reference files.

Source summary

• feat(cli): add sandbox sessions and agents subcommand groups #4615 and feat(onboard): bake secondary agents via NEMOCLAW_EXTRA_AGENTS_JSON #4653 -> docs/about/release-notes.mdx, docs/reference/commands.mdx: Release notes now cover host-side sessions and agents commands plus NEMOCLAW_EXTRA_AGENTS_JSON secondary-agent baking.
• fix(docker): replace docker pull hard timeout with progress watchdog for vllm #4163, fix(inference): request streaming usage for local ollama #4204, fix(connect): stop silent gateway inference revert #4611, feat(inference): switches the DGX Spark managed-vLLM profile to Qwen3.6-35B-A3B-NVFP4 #4619, and fix(inference): improve managed vLLM download and launch progress #4676 -> docs/about/release-notes.mdx, docs/inference/use-local-inference.mdx: Release notes now cover managed vLLM progress/readiness, DGX Spark model default changes, local Ollama streaming usage, and inference route divergence warnings.
• feat(onboard): auto-apply UFW sandbox-bridge rule under NEMOCLAW_AUTO_FIX_FIREWALL=1 (#4265) #4267, fix(onboard): reuse containerized gateway and repair routed provider reachability (#4520, #4564) #4601, fix(onboard): gate host-network GPU local inference reachability (#4509) #4609, fix(onboard): don't leave a default sandbox when cancelling at policy presets #4642, fix(onboard): bind gateway name, state dir, and marker per gateway port (#4422) #4645, and fix(onboard): preserve effective policy preset selection across re-onboard #4661 -> docs/about/release-notes.mdx, docs/reference/commands.mdx: Release notes now cover UFW auto-remediation, local-inference reachability gates, gateway reuse/binding, cancel rollback, and policy selection persistence.
• fix(cli): make channels add atomic when policy preset is missing #4577, fix: validate Slack credentials before enabling channels #4582, fix(whatsapp): render compact pairing QR and split gateway diagnostics #4607, and fix(channels): normalize Slack runtime env placeholders before OpenClaw start (#4274) #4660 -> docs/about/release-notes.mdx, docs/manage-sandboxes/messaging-channels.mdx: Release notes now cover Slack validation, atomic channels add, WhatsApp QR diagnostics, and Slack placeholder normalization.
• fix(cli): surface docker/sandbox container/dashboard port failure layers in sandbox status #4388, fix(sandbox): stop false Docker unhealthy and add paused-container hint (#4503, #4495) #4600, fix(cli): skip stale k3s gateway container check in Docker-driver doctor (#4502) #4646, and fix(sandbox): preserve registry entry on stale-sandbox connect recovery (#4497) #4647 -> docs/about/release-notes.mdx, docs/reference/commands.mdx: Release notes now cover status failure layers, paused-container hints, Docker-driver doctor behavior, and non-destructive stale-registry recovery.
• fix(policy): allow uv in pypi preset #4569, fix(policy): make Jira curl validation observable #4579, and fix(installer): clarify install tag pinning #4678 -> docs/about/release-notes.mdx, docs/manage-sandboxes/lifecycle.mdx, docs/network-policy/integration-policy-examples.mdx: Release notes now cover installer tag pinning, PyPI uv policy access, and observable Jira validation.
• docs: add per-agent documentation variants #4632 -> .agents/skills/: Regenerated user skills from the current per-agent docs source, including newly generated Hermes reference files.

Verification

• python3 scripts/docs-to-skills.py docs/ .agents/skills/ --prefix nemoclaw-user --doc-platform fern-mdx
• rg "permissive mode|shields down|shields up|shields status|config rotate-token|rotate-token" docs --glob "*.mdx"
• rg "permissive mode|shields down|shields up|shields status|config rotate-token|rotate-token" .agents/skills --glob "*.md"
• npm run docs
• npm run build:cli
• Commit hooks: markdownlint, docs-to-skills verification, gitleaks, skills YAML, commitlint

Summary by CodeRabbit

• Documentation
  • Restructured documentation to clearly distinguish OpenClaw and Hermes agent variants throughout user guides.
  • Enhanced security, credential storage, and deployment guidance with clearer setup flows.
  • Added Hermes plugin installation and ecosystem documentation.
  • Improved workspace, messaging, and policy management references with variant-specific command examples.
  • Refined troubleshooting and CLI reference sections for clarity.

[coderabbitai]

Caution

Review failed

Pull request was closed or merged during review

📝 Walkthrough

Walkthrough

This PR introduces comprehensive documentation refactoring across the NemoClaw skill guides and reference materials to enable OpenClaw vs Hermes variant-specific rendering via AgentOnly template blocks. Major changes include restructured inference configuration guidance, expanded security documentation with Hermes-specific sections, updated sandbox lifecycle and policy management documentation, standardized code fence formatting from console to bash style, and v0.0.57 release notes.

Changes

Ecosystem, Architecture, and Overview Documentation

Layer / File(s)

Summary

Overview, architecture, and ecosystem documentation .agents/skills/nemoclaw-user-overview/SKILL.md, references/ecosystem-hermes.md, references/ecosystem.md, references/how-it-works.md, references/overview.md, .agents/skills/nemoclaw-user-reference/SKILL.md, references/architecture.md

High-level documentation restructured to use AgentCli/AgentOnly components instead of hardcoded CLI names. New ecosystem-hermes.md page added describing Hermes/OpenShell/NemoClaw stack relationships and onboarding comparison. Architecture documentation updated to describe NemoClaw as a generic integration layer over compatible agents (OpenClaw and Hermes) rather than OpenClaw-specific plugin orchestration. Mermaid diagrams and variant-specific Next Steps sections added.

Inference Configuration and Provider Setup

Layer / File(s)	Summary
Inference configuration AgentOnly refactoring .agents/skills/nemoclaw-user-configure-inference/SKILL.md, references/inference-options.md, references/switch-inference-providers.md, references/tool-calling-reliability.md, references/set-up-sub-agent.md	Comprehensive update adding AgentOnly blocks to split OpenClaw vs Hermes examples for onboarding, provider selection, Local Ollama setup (with token/authenticated-proxy behavior), vLLM/Anthropic/NIM configuration, and runtime model switching. Tool-calling reliability guide clarified with structured tool_calls requirement emphasis. Sub-agent setup converted to bash fence formatting.

Security and Credential Management

Layer / File(s)

Summary

Security best practices and credential storage .agents/skills/nemoclaw-user-configure-security/references/best-practices.md, references/credential-storage.md, references/openclaw-controls.md

Security documentation expanded with OpenClaw variant-specific config directory writability posture, new Hermes directory section describing /sandbox/.hermes/ state paths and regeneration behavior, and new claude-code preset warning clarifying it is not for inference routing. Operator approval flow refined to describe durable policy revisions persisting across restarts within a sandbox instance. Credential storage refactored into OpenClaw vs Hermes AgentOnly blocks showing variant-specific onboarding and credential rotation commands. GitHub token and keychain equivalence clarified.

Sandbox Management and Day-2 Operations

Layer / File(s)

Summary

Sandbox lifecycle and operational management .agents/skills/nemoclaw-user-manage-sandboxes/SKILL.md, references/backup-restore.md, references/install-plugins-hermes.md, references/messaging-channels.md, references/runtime-controls.md, references/workspace-files.md, .agents/skills/nemoclaw-user-monitor-sandbox/SKILL.md

nemoclaw-user-manage-sandboxes SKILL.md comprehensively updated with OpenClaw vs Hermes command variants for listing, status checking, logging, diagnostics, rebuilding, and uninstall operations. New install-plugins-hermes.md documents Hermes plugin installation via custom Dockerfile approach (not skill install). backup-restore.md, messaging-channels.md, runtime-controls.md, and workspace-files.md split into variant-specific AgentOnly blocks showing different state locations, mutability timing rules, and management commands per agent. Monitor-sandbox SKILL.md updated with variant-specific commands for status, logs, and test-inference operations.

Policy and Network Management

Layer / File(s)

Summary

Policy management and network configuration .agents/skills/nemoclaw-user-manage-policy/SKILL.md, references/approve-network-requests.md, references/integration-policy-examples.md, .agents/skills/nemoclaw-user-reference/references/network-policies.md

nemoclaw-user-manage-policy SKILL.md updated with AgentOnly blocks splitting all policy commands (policy-add, policy-list, policy-remove, rebuild) between nemoclaw and nemohermes variants. approve-network-requests.md and integration-policy-examples.md converted to bash code fences with AgentOnly blocks for all provider and channel management examples. network-policies.md clarified with explicit statement that claude-code direct egress is not included in any policy tier and requires explicit preset application.

Getting Started and Remote Deployment

Layer / File(s)

Summary

Getting started onboarding and remote deployment .agents/skills/nemoclaw-user-get-started/SKILL.md, references/prerequisites.md, references/quickstart-hermes.md, references/windows-preparation.md, .agents/skills/nemoclaw-user-deploy-remote/SKILL.md, references/brev-web-ui.md, references/install-openclaw-plugins.md, references/sandbox-hardening.md

nemoclaw-user-get-started SKILL.md refined with prerequisites review guidance, third-party software acceptance for non-TTY contexts, and clearer Local Ollama/vLLM detection behavior. windows-preparation.md restructured into AgentOnly blocks for openclaw vs hermes variant-specific instructions. nemoclaw-user-deploy-remote SKILL.md reorganized with explicit command blocks for remote TUI entry, test inference, CHAT_UI_URL timing, sandbox readiness budget (NEMOCLAW_SANDBOX_READY_TIMEOUT), and proxy configuration. Supporting references updated with bash fence standardization and improved formatting.

CLI Reference and Troubleshooting

Layer / File(s)

Summary

Commands reference and troubleshooting consolidation .agents/skills/nemoclaw-user-reference/references/commands.md, references/cli-selection-guide.md, references/troubleshooting.md

commands.md massively expanded with AgentOnly-gated documentation across all nemoclaw/nemohermes subcommands, new NEMOCLAW_AUTO_FIX_FIREWALL variable, expanded status/failure-layer descriptions, environment variable schema tables, and onboarding configuration rows. cli-selection-guide.md converted to bash code fences. troubleshooting.md extensively reformatted with 60+ bash fence standardizations across all command examples (console → bash with $ prompts removed) while preserving underlying diagnostic logic and instructions.

Release Notes and Documentation Formatting

Layer / File(s)

Summary

Release v0.0.57 notes and MDX formatting docs/about/release-notes.mdx, docs/inference/switch-inference-providers.mdx, docs/manage-sandboxes/lifecycle.mdx, docs/manage-sandboxes/messaging-channels.mdx, docs/reference/architecture.mdx, docs/reference/network-policies.mdx

New v0.0.57 release section added with highlights across multi-agent CLI workflows, local inference setup observability, messaging channel reliability, sandbox diagnostics output, and installer/policy tightening. MDX files receive whitespace/formatting adjustments for readability and clarifications on policy tiers (claude-code explicit opt-in) and Hermes policy defaults (only policy-additions.yaml, no permissive variant).

---

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~25 minutes

Possibly related PRs

• NVIDIA/NemoClaw#4632: Introduces AgentGuide variant-aware routing infrastructure that underpins the AgentOnly blocks used throughout this PR.
• NVIDIA/NemoClaw#4075: Adds claude-code policy preset and removes it from permissive policies, directly referenced by the new claude-code security warnings in this PR.
• NVIDIA/NemoClaw#4618: Earlier Model Router documentation changes that are extended with variant-specific blocks in this PR's inference-options.md refactor.

Suggested reviewers

• cv
• ericksoa
• prekshivyas

Poem

🐇 A rabbit hopped through docs with glee,
Agent-aware, now Hermes and Claw you'll see!
Bash fences gleam, variant blocks align,
Blue wizard ways refactored fine—
One path, two agents, seamlessly divine! 🌟

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch docs/0.0.57-refresh

[github-actions]

🌿 Preview your docs: https://nvidia-preview-pr-4716.docs.buildwithfern.com/nemoclaw

[github-actions]

E2E Advisor Recommendation

Required E2E: None
Optional E2E: None

Workflow run

Full advisor summary

E2E Recommendation Advisor

Base: origin/main
Head: HEAD
Confidence: high

Required E2E

• None. No E2E is recommended because this PR only changes documentation and agent-skill Markdown content. Although the topics cover inference, policy, security, deployment, and sandbox lifecycle, no runtime code, installer logic, credentials handling, blueprint assets, network-policy YAML, or real assistant execution paths are changed. Existing docs validation/link/preview checks are the appropriate CI coverage, not sandbox E2E.

Optional E2E

• None.

New E2E recommendations

• None.

[github-actions]

E2E Scenario Advisor Recommendation

Required scenario E2E: None
Optional scenario E2E: None

Workflow run

Full scenario advisor summary

E2E Scenario Advisor

Base: origin/main
Head: HEAD
Confidence: high

Required scenario E2E

• None. Only documentation and generated user agent skill Markdown files changed; no scenario runtime, workflow, metadata, expected-state, suite, onboarding helper, or scenario-relevant implementation files were modified.

Optional scenario E2E

• None.

Relevant changed files

• None.

[github-actions]

PR Review Advisor

Findings: 1 needs attention, 1 worth checking, 0 nice ideas
Top item: Fix malformed local inference generated skill

Review findings

🛠️ Needs attention

• Regenerated local-inference skill is malformed (.agents/skills/nemoclaw-user-configure-inference/SKILL.md:17): The generated skill no longer matches the source MDX structure. The `openclaw` AgentOnly block starts under prerequisites, includes both the OpenClaw and Hermes prerequisite bullets, then the MDX `import { AgentOnly } ...` and much of the page body appear inside that still-open block. This incorrectly scopes the local-inference skill content for agent consumers and leaves the Hermes prerequisite outside its intended variant block.
  • Recommendation: Regenerate or fix the docs-to-skills conversion so the skill preserves the source order from `docs/inference/use-local-inference.mdx`: top-level import/intro first, then separate closed OpenClaw and Hermes prerequisite blocks, followed by the shared local-inference content.
  • Evidence: Source `docs/inference/use-local-inference.mdx` has the import and intro before `## Prerequisites`, with separate `<AgentOnly variant="openclaw">` and `<AgentOnly variant="hermes">` blocks. The generated skill at lines 17-25 opens only the OpenClaw block, includes both prerequisite bullets, and places the import inside that block before shared content.

🔎 Worth checking

• Hermes command reference drops the Hermes installer selector (.agents/skills/nemoclaw-user-reference/references/commands-nemohermes.md:106): The NemoHermes command reference's manual OpenShell upgrade recovery path tells users to rerun the installer as `curl -fsSL https://www.nvidia.com/nemoclaw.sh | NEMOCLAW_OPENSHELL_UPGRADE_PREPARED=1 bash`. In the Hermes lifecycle docs, the corresponding command includes `NEMOCLAW_AGENT=hermes`. Without that selector, a user following the Hermes-specific generated reference may re-enter the default OpenClaw installer path after doing Hermes-specific prep.
  • Recommendation: Update the generated Hermes command reference, or its source transformation, so this inline installer command includes `NEMOCLAW_AGENT=hermes NEMOCLAW_OPENSHELL_UPGRADE_PREPARED=1 bash` like the Hermes lifecycle guidance.
  • Evidence: `docs/manage-sandboxes/lifecycle.mdx` Hermes guidance includes `curl -fsSL https://www.nvidia.com/nemoclaw.sh | NEMOCLAW_AGENT=hermes NEMOCLAW_OPENSHELL_UPGRADE_PREPARED=1 bash`, while `.agents/skills/nemoclaw-user-reference/references/commands-nemohermes.md` line 106 omits `NEMOCLAW_AGENT=hermes`.

🌱 Nice ideas

• None.

Workflow run details

This is an automated advisory review. A human maintainer must make the final merge decision.