docs(skills): make copy of docs-as-skills for Claude - #837
Merged
Conversation
Contributor
|
Caution Review failedPull request was closed or merged during review 📝 WalkthroughWalkthroughThis PR switches NemoClaw inference from NVIDIA-specific routing to a routed, multi-provider model via OpenShell, adds runtime model-switching docs and workspace documentation, refactors the docs generator to support multiple output directories and create a Changes
Sequence Diagram(s)sequenceDiagram
autonumber
participant User as User CLI
participant OpenShell as OpenShell (host)
participant Sandbox as Sandbox (inference.local)
participant Provider as Upstream Provider
User->>OpenShell: runs `openshell inference set --provider <P> --model <M>`
OpenShell->>OpenShell: update routing configuration for `inference.local` -> <P>/<M>
Note over OpenShell,Sandbox: Sandbox keeps using `inference.local` (no restart)
Sandbox->>OpenShell: POST /v1/... (inference request to `inference.local`)
OpenShell->>Provider: forward request to configured provider endpoint
Provider-->>OpenShell: inference response
OpenShell-->>Sandbox: return inference response
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~25 minutes Poem
🚥 Pre-merge checks | ✅ 2 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (2 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
Comment |
Collaborator
|
Why not a symlink instead of duplicating all these files? |
Collaborator
Author
I thought about it as well. I'm open to that approach. |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
Contributor
There was a problem hiding this comment.
Actionable comments posted: 19
🧹 Nitpick comments (2)
.agents/skills/docs/nemoclaw-overview/references/how-it-works.md (1)
96-97: Clarify interplay between onboarding bake-in and runtime model switching.Lines 96-97 are directionally correct, but they can read as immutable after image creation. Consider adding one sentence that runtime
openshell inference setcan still change active routing without rebuilding the sandbox.🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In @.agents/skills/docs/nemoclaw-overview/references/how-it-works.md around lines 96 - 97, Clarify that the model reference baked into the sandbox during onboarding is not immutable by adding a sentence after the sentence that mentions "bakes the matching model reference into the sandbox image" explaining that the runtime OpenShell command openshell inference set can reconfigure the active routing to a different provider/model (changing where inference.local points) without rebuilding the sandbox image; reference "inference.local", "sandbox image", and the CLI command "openshell inference set" so readers know how to perform runtime switching..claude/skills/nemoclaw-manage-policy/SKILL.md (1)
3-3: Tighten the skill metadata description grammar and phrasing.The current description is grammatically inconsistent (“Reviews and approve”, “covers adds, remove”) and overly keyword-stacked, which reduces clarity in skill discovery.
✍️ Suggested frontmatter text
-description: Reviews and approve blocked agent network requests in the TUI. Also covers adds, remove, or modify allowed endpoints in the sandbox policy. Use when approve deny nemoclaw agent, customize nemoclaw network policy, customize nemoclaw sandbox network, nemoclaw, nemoclaw approve network requests, network policy, openclaw, openshell. +description: Review and approve blocked agent network requests in the TUI. Also covers how to add, remove, or modify allowed endpoints in the sandbox policy. Use this skill for NemoClaw network-policy and OpenShell approval workflows.🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In @.claude/skills/nemoclaw-manage-policy/SKILL.md at line 3, Update the frontmatter "description" value to a concise, grammatically correct sentence that describes the skill's purpose and removes redundant keywords; specifically replace the current text under the description key (the string that begins "Reviews and approve blocked agent network requests...") with a clear phrase such as: "Review and approve or deny blocked agent network requests in the TUI, and add, remove, or modify allowed endpoints in the sandbox network policy." Ensure you only change the description field content and keep other metadata intact.
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In @.agents/skills/docs/nemoclaw-reference/references/inference-profiles.md:
- Line 64: Replace the plain text "Switch Inference Models" with an explicit
relative markdown link pointing to the corresponding skill page so readers can
click through directly; locate the sentence containing "Switch Inference Models
(see the `nemoclaw-configure-inference` skill)" and update it to use a markdown
link referencing the "nemoclaw-configure-inference" skill page (i.e., link the
"Switch Inference Models" text to that skill's relative URL).
In @.agents/skills/docs/nemoclaw-reference/SKILL.md:
- Line 3: Fix the typo in the SKILL.md metadata description: replace "fulls CLI
reference" with "full CLI reference" (or "full CLI command reference") in the
description field so it reads correctly; update the description string in the
metadata block for the NemoClaw skill (the description property currently
containing "fulls CLI reference") accordingly.
In @.agents/skills/docs/nemoclaw-workspace/references/workspace-files.md:
- Around line 47-49: The sentence in the warning is self-referential: replace
"See Back Up and Restore (see the `nemoclaw-workspace` skill) for instructions."
with an in-page/in-skill reference that points directly to the Back Up and
Restore section within this document or skill (e.g., "See the Back Up and
Restore section below for instructions." or "Refer to the 'Back Up and Restore'
section in this skill for instructions."), updating the text near the Warning in
workspace-files.md to remove the back-reference to `nemoclaw-workspace` and
instead link to the internal section heading or anchor for Back Up and Restore.
In @.agents/skills/docs/nemoclaw-workspace/SKILL.md:
- Line 3: The frontmatter "description" value in SKILL.md contains typos and
awkward phrasing; update the description key to a clear, grammatical sentence
such as: "How to back up and restore OpenClaw workspace files before destructive
operations; explains what workspace files are, where they reside, and how they
persist across sandbox restarts." Locate the "description" frontmatter field in
.agents/skills/docs/nemoclaw-workspace/SKILL.md and replace the existing text
with the corrected sentence, keeping it concise and using proper punctuation.
In @.claude/skills/nemoclaw-configure-inference/SKILL.md:
- Around line 23-57: Update the SKILL.md examples to make it explicit that the
provider IDs shown (nvidia-prod, openai-api, anthropic-prod, gemini-api,
compatible-endpoint, compatible-anthropic-endpoint) are illustrative and not
universal defaults; add a short note near the "NVIDIA
Endpoints"/"OpenAI"/"Anthropic"/"Google Gemini"/"Compatible Endpoints" blocks
stating users should substitute the provider ID used in their deployment and
verify available provider IDs (e.g. via the openshell provider list command)
before running the openshell inference set --provider ... --model ... commands.
In @.claude/skills/nemoclaw-deploy-remote/SKILL.md:
- Line 15: The prerequisite line stating "A running NemoClaw sandbox, either
local or remote." is incorrect; remove that sentence from SKILL.md and update
the prerequisites section to state that "nemoclaw deploy" will provision and
launch a sandbox for first-time remote deployments (or add a short note
clarifying that an existing sandbox is only needed for iterative deploys),
ensuring the guide no longer requires a prior running sandbox.
In @.claude/skills/nemoclaw-manage-policy/SKILL.md:
- Line 132: Replace the phrase "re-run setup" with the consistent command
reference used elsewhere (specifically "nemoclaw onboard") so the sentence reads
that to make changes permanent, update the static policy file and run `nemoclaw
onboard`; ensure this matches the wording used in Step 5 and any other places
that reference applying static policy updates.
In @.claude/skills/nemoclaw-monitor-sandbox/SKILL.md:
- Around line 69-72: Step 4 uses inconsistent sandbox placeholders: one command
shows "nemoclaw my-assistant connect" while the rest uses "<name>"; update the
commands so they use the same placeholder (e.g., replace "my-assistant" with
"<name>" or vice versa) so both lines consistently reference the same sandbox
identifier (refer to the commands "nemoclaw my-assistant connect" and "openclaw
agent --agent main --local -m \"Test inference\" --session-id debug" when making
the change).
In @.claude/skills/nemoclaw-overview/references/how-it-works.md:
- Line 26: Replace the provider-specific node label "INF[NVIDIA inference,
routed]" with a provider-agnostic label in the Mermaid diagram so routing is
generic; locate the INF[...] token in
.claude/skills/nemoclaw-overview/references/how-it-works.md and change the text
inside the brackets to something like "Inference provider, routed" or "Inference
(routed)" to reflect a generic routed-provider model.
In @.claude/skills/nemoclaw-overview/references/overview.md:
- Around line 8-13: Update the compound adjective usages of "open source" to
"open-source" in the overview text: change "open source reference stack" to
"open-source reference stack" and "open source models" to "open-source models"
(leave standalone uses unchanged if any); edit the sentences mentioning "open
source" alongside "models" and "reference stack" in the NemoClaw overview so
hyphenation is consistent across the paragraph.
In @.claude/skills/nemoclaw-overview/SKILL.md:
- Line 3: The frontmatter description field contains grammatical errors and
awkward phrases; rewrite the description value for clarity and discoverability
by correcting capitalization and pluralization (e.g., "NemoClaw", "OpenClaw"),
removing incorrect fragments like "overview does fits" and "nemoClaws is", and
producing a concise, searchable sentence that mentions the CLI plugin, versioned
blueprint, sandboxing OpenClaw, and that NemoClaw is an open-source reference
stack with changelogs and release notes; update the description string in the
"description:" frontmatter entry to this cleaned, keyword-preserving sentence.
- Line 35: The diagram node labeled INF[NVIDIA inference, routed] incorrectly
implies NVIDIA-only routing; update that node (INF[...] in SKILL.md) to a
provider-neutral label such as INF[Inference, routed] or INF[Multi-provider
inference, routed] and ensure any adjacent labels or comments reflect
multi-provider routing (OpenAI, Anthropic, Gemini, compatible endpoints) so the
flowchart matches the PR docs.
In @.claude/skills/nemoclaw-reference/references/architecture.md:
- Around line 75-80: The diagram in the "Inference requests from the agent..."
paragraph hardcodes "NVIDIA Endpoint (build.nvidia.com)" which conflicts with
the multi-provider routing described elsewhere; update the ASCII diagram to use
a generic provider name (e.g., "Inference Provider Endpoint" or "Configured
Provider") or show multiple provider options and remove the NVIDIA-specific URL,
keeping the rest of the wording intact and preserving "Agent (sandbox)" and
"OpenShell gateway" labels to make multi-provider routing clear.
In @.claude/skills/nemoclaw-reference/references/commands.md:
- Around line 84-85: Replace the legacy provider-specific phrase "Stop the NIM
container and delete the sandbox." in the destroy command docs with
provider-agnostic wording such as "Stop inference services and delete the
sandbox" and similarly update the following line "This removes the sandbox from
the registry." if needed to match neutral language; locate the string in the
destroy command documentation (the lines containing "Stop the NIM container..."
and "This removes the sandbox from the registry.") and swap in the suggested
neutral phrasing so the docs reflect the routed multi-provider model.
In @.claude/skills/nemoclaw-reference/references/network-policies.md:
- Around line 25-79: The document currently uses a MyST list-table directive
(`:::{list-table}`) and needs to be converted to a plain Markdown table: replace
the entire directive block (the header and rows containing policies like
`claude_code`, `nvidia`, `github`, `github_rest_api`, `clawhub`, `openclaw_api`,
`openclaw_docs`, `npm_registry`, `telegram`) with a standard pipe-delimited
Markdown table that preserves the four columns ("Policy", "Endpoints",
"Binaries", "Rules") and keeps backtick formatting for code elements (e.g.,
`api.anthropic.com:443`, `/usr/local/bin/claude`, `github.com:443`, etc.),
ensuring each policy is a separate row and the table header and separator row
are present so Markdown renderers treat it as a table.
In @.claude/skills/nemoclaw-reference/references/troubleshooting.md:
- Around line 19-20: The sentence "NemoClaw requires Linux Ubuntu 22.04 LTS or
later" is too broad and conflicts with later macOS troubleshooting; update the
troubleshooting doc to explicitly scope the platform requirement by clarifying
that the core runtime/sandbox host requires Linux (e.g., "Sandbox
host/production nodes require Ubuntu 22.04 LTS or later"), while operator
workstations may be macOS or Windows using Docker Desktop/Colima for local
development; mention the phrases "NemoClaw requires Linux Ubuntu 22.04 LTS or
later" and the macOS troubleshooting sections so you add a short clarifying
sentence under that line and/or an explicit subheading (e.g., "Platform scope:
sandbox host vs operator workstation") that directs users to the macOS
Docker/Colima guidance for developer machines.
In @.claude/skills/nemoclaw-reference/SKILL.md:
- Line 3: The skill description string in SKILL.md contains a typo ("fulls CLI
reference"); update the description value to read "full CLI reference" (or "full
CLI command reference") so the sentence reads correctly, e.g., replace "fulls
CLI reference for plugin and standalone NemoClaw commands" with "full CLI
reference for plugin and standalone NemoClaw commands" in the description field.
In @.claude/skills/nemoclaw-workspace/SKILL.md:
- Line 3: Update the frontmatter "description" value to a grammatically correct,
concise sentence: replace the current run-on text with something like "How to
back up and restore OpenClaw workspace files before destructive operations,
including what workspace files are, where they live, and how they persist across
sandbox restarts." Ensure you edit the frontmatter key named description in
SKILL.md to use this corrected phrasing so search/discovery metadata is clear.
In `@scripts/docs-to-skills.py`:
- Around line 43-47: Update the CLI examples in scripts/docs-to-skills.py to use
the correct output directory (.agents/skills/docs/) instead of .agents/skills/
so generated files land in the intended docs subfolder; search for the example
invocations (the lines showing "python scripts/docs-to-skills.py docs/
.agents/skills/ ..." and similar repeats) and change those occurrences to
".agents/skills/docs/" (also update any related examples that combine
--strategy, --prefix, --dry-run, --name-map, or --exclude flags).
---
Nitpick comments:
In @.agents/skills/docs/nemoclaw-overview/references/how-it-works.md:
- Around line 96-97: Clarify that the model reference baked into the sandbox
during onboarding is not immutable by adding a sentence after the sentence that
mentions "bakes the matching model reference into the sandbox image" explaining
that the runtime OpenShell command openshell inference set can reconfigure the
active routing to a different provider/model (changing where inference.local
points) without rebuilding the sandbox image; reference "inference.local",
"sandbox image", and the CLI command "openshell inference set" so readers know
how to perform runtime switching.
In @.claude/skills/nemoclaw-manage-policy/SKILL.md:
- Line 3: Update the frontmatter "description" value to a concise, grammatically
correct sentence that describes the skill's purpose and removes redundant
keywords; specifically replace the current text under the description key (the
string that begins "Reviews and approve blocked agent network requests...") with
a clear phrase such as: "Review and approve or deny blocked agent network
requests in the TUI, and add, remove, or modify allowed endpoints in the sandbox
network policy." Ensure you only change the description field content and keep
other metadata intact.
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
Run ID: cb51e337-4767-4210-892b-76e4ff65ab9f
📒 Files selected for processing (28)
.agents/skills/docs/nemoclaw-configure-inference/SKILL.md.agents/skills/docs/nemoclaw-get-started/SKILL.md.agents/skills/docs/nemoclaw-overview/references/how-it-works.md.agents/skills/docs/nemoclaw-reference/SKILL.md.agents/skills/docs/nemoclaw-reference/references/commands.md.agents/skills/docs/nemoclaw-reference/references/inference-profiles.md.agents/skills/docs/nemoclaw-reference/references/troubleshooting.md.agents/skills/docs/nemoclaw-workspace/SKILL.md.agents/skills/docs/nemoclaw-workspace/references/workspace-files.md.claude/skills/nemoclaw-configure-inference/SKILL.md.claude/skills/nemoclaw-deploy-remote/SKILL.md.claude/skills/nemoclaw-get-started/SKILL.md.claude/skills/nemoclaw-manage-policy/SKILL.md.claude/skills/nemoclaw-monitor-sandbox/SKILL.md.claude/skills/nemoclaw-overview/SKILL.md.claude/skills/nemoclaw-overview/references/how-it-works.md.claude/skills/nemoclaw-overview/references/overview.md.claude/skills/nemoclaw-overview/references/release-notes.md.claude/skills/nemoclaw-reference/SKILL.md.claude/skills/nemoclaw-reference/references/architecture.md.claude/skills/nemoclaw-reference/references/commands.md.claude/skills/nemoclaw-reference/references/inference-profiles.md.claude/skills/nemoclaw-reference/references/network-policies.md.claude/skills/nemoclaw-reference/references/troubleshooting.md.claude/skills/nemoclaw-workspace/SKILL.md.claude/skills/nemoclaw-workspace/references/workspace-files.md.gitignorescripts/docs-to-skills.py
💤 Files with no reviewable changes (1)
- .gitignore
Contributor
There was a problem hiding this comment.
♻️ Duplicate comments (1)
.agents/skills/nemoclaw-reference/references/inference-profiles.md (1)
64-64:⚠️ Potential issue | 🟡 MinorAdd a clickable link for “Switch Inference Models.”
At Line 64, the reference is still plain text, which adds friction for navigation. Please link it directly to the relevant skill page.
Suggested fix
-For runtime switching guidance, refer to Switch Inference Models (see the `nemoclaw-configure-inference` skill). +For runtime switching guidance, refer to [Switch Inference Models](../../nemoclaw-configure-inference/SKILL.md) (see the `nemoclaw-configure-inference` skill).🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In @.agents/skills/nemoclaw-reference/references/inference-profiles.md at line 64, Replace the plain text "Switch Inference Models" with a clickable link to the skill page by editing the sentence that currently reads "For runtime switching guidance, refer to Switch Inference Models (see the `nemoclaw-configure-inference` skill)." — update the phrase "Switch Inference Models" to a Markdown link pointing to the corresponding skill document (the `nemoclaw-configure-inference` skill page) so readers can click directly to the referenced skill.
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Duplicate comments:
In @.agents/skills/nemoclaw-reference/references/inference-profiles.md:
- Line 64: Replace the plain text "Switch Inference Models" with a clickable
link to the skill page by editing the sentence that currently reads "For runtime
switching guidance, refer to Switch Inference Models (see the
`nemoclaw-configure-inference` skill)." — update the phrase "Switch Inference
Models" to a Markdown link pointing to the corresponding skill document (the
`nemoclaw-configure-inference` skill page) so readers can click directly to the
referenced skill.
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
Run ID: cd8064b7-69d0-4559-89d7-781803ba34e3
📒 Files selected for processing (19)
.agents/skills/nemoclaw-configure-inference/SKILL.md.agents/skills/nemoclaw-deploy-remote/SKILL.md.agents/skills/nemoclaw-get-started/SKILL.md.agents/skills/nemoclaw-manage-policy/SKILL.md.agents/skills/nemoclaw-monitor-sandbox/SKILL.md.agents/skills/nemoclaw-overview/SKILL.md.agents/skills/nemoclaw-overview/references/how-it-works.md.agents/skills/nemoclaw-overview/references/overview.md.agents/skills/nemoclaw-overview/references/release-notes.md.agents/skills/nemoclaw-reference/SKILL.md.agents/skills/nemoclaw-reference/references/architecture.md.agents/skills/nemoclaw-reference/references/commands.md.agents/skills/nemoclaw-reference/references/inference-profiles.md.agents/skills/nemoclaw-reference/references/network-policies.md.agents/skills/nemoclaw-reference/references/troubleshooting.md.agents/skills/nemoclaw-workspace/SKILL.md.agents/skills/nemoclaw-workspace/references/workspace-files.md.claude/skillsscripts/docs-to-skills.py
✅ Files skipped from review due to trivial changes (8)
- .agents/skills/nemoclaw-reference/references/troubleshooting.md
- .agents/skills/nemoclaw-reference/SKILL.md
- .agents/skills/nemoclaw-configure-inference/SKILL.md
- .agents/skills/nemoclaw-workspace/references/workspace-files.md
- .claude/skills
- .agents/skills/nemoclaw-workspace/SKILL.md
- .agents/skills/nemoclaw-get-started/SKILL.md
- .agents/skills/nemoclaw-reference/references/commands.md
🚧 Files skipped from review as they are similar to previous changes (1)
- scripts/docs-to-skills.py
miyoungc
changed the title
cv
approved these changes
Mar 24, 2026
temrjan
pushed a commit
to temrjan/NemoClaw
that referenced
this pull request
Mar 25, 2026
* docs: docs-to-skills for claude * docs: remove wrong target skills and rerun the conversion * docs: use symlink * fix: run pre-commit * docs: pre-commit --------- Co-authored-by: Carlos Villela <cvillela@nvidia.com>
This was referenced Apr 2, 2026
lakamsani
pushed a commit
to lakamsani/NemoClaw
that referenced
this pull request
Apr 4, 2026
* docs: docs-to-skills for claude * docs: remove wrong target skills and rerun the conversion * docs: use symlink * fix: run pre-commit * docs: pre-commit --------- Co-authored-by: Carlos Villela <cvillela@nvidia.com>
16 tasks
jacobtomlinson
pushed a commit
to jacobtomlinson/NemoClaw
that referenced
this pull request
Apr 30, 2026
* docs: docs-to-skills for claude * docs: remove wrong target skills and rerun the conversion * docs: use symlink * fix: run pre-commit * docs: pre-commit --------- Co-authored-by: Carlos Villela <cvillela@nvidia.com>
wscurran
added
area: docs
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Related Issue
Changes
Type of Change
Testing
npx prek run --all-filespasses (or equivalentlymake check).npm testpasses.make docsbuilds without warnings. (for doc-only changes)Checklist
General
Code Changes
npx prek run --all-filesauto-fixes formatting (ormake formatfor targeted runs).Doc Changes
update-docsagent skill to draft changes while complying with the style guide. For example, prompt your agent with "/update-docscatch up the docs for the new changes I made in this PR."Summary by CodeRabbit
New Features
Documentation