Skip to content
Merged
Show file tree
Hide file tree
Changes from 1 commit
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1 change: 1 addition & 0 deletions .agents/skills/nemoclaw-user-configure-inference/SKILL.md
Original file line number Diff line number Diff line change
Expand Up @@ -73,6 +73,7 @@ NemoClaw lists installed models or offers starter models if none are installed.
On hosts where the larger starter models fit the currently available GPU memory, the starter list includes `qwen3.6:35b` and selects it by default.
When another GPU workload is using most of the memory at onboard time, NemoClaw downgrades the menu to the largest model that still fits.
It pulls the selected model, loads it into memory, and validates it before continuing.
When Ollama reports a loaded-model context length, NemoClaw uses that value for the `contextWindow` baked into `openclaw.json` unless you set `NEMOCLAW_CONTEXT_WINDOW` yourself.
If the selected model declares that it does not support tool calling, onboarding stops with guidance to choose a model whose `ollama show <model>` capabilities include `tools`.
The validation also requires structured chat-completions tool calls.
If the model leaks tool-call JSON as plain message text, onboarding stops so you can choose a model that returns tool calls in the expected response field.
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -123,6 +123,7 @@ To change these values, set the corresponding environment variables before runni
| `NEMOCLAW_AGENT_HEARTBEAT_EVERY` | Go-style duration (`30m`, `1h`, `0m` to disable) | `unset` (OpenClaw default) |

Invalid values are ignored, and the default bakes into the image.
For Local Ollama, onboarding loads the selected model first and uses Ollama's reported runtime context length when `NEMOCLAW_CONTEXT_WINDOW` is unset.
Use `NEMOCLAW_INFERENCE_INPUTS=text,image` only for a model that accepts image input through the selected provider.

```console
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -182,13 +182,13 @@ Writable agent state such as plugins, skills, hooks, and workspace metadata live
By default, this directory starts writable so the agent can manage its own config, install skills, and write to standard home-directory paths natively.
For sensitive workloads, use a reviewed host-side immutability workflow after initial setup so config and writable state entry points cannot be changed by the sandbox user.

- **DAC permissions (default).** The sandbox user owns `/sandbox/.openclaw` with mode `700` and `openclaw.json` with mode `600`, so the agent can read and write config directly.
- **DAC permissions (default).** The sandbox user owns `/sandbox/.openclaw` with mode `2770` (setgid `sandbox:sandbox`) and `openclaw.json` with mode `660`, so the agent and its group can read and write config directly. A reviewed host-side immutability workflow should compare the intended ownership and mode with the live sandbox filesystem before treating the config tree as locked.
- **Config integrity hash.** The image includes a SHA256 hash of `openclaw.json`. In the default mutable state, `.config-hash` is sandbox-owned and is not a tamper-proof trust anchor, so startup does not fail closed on that hash. When the hash is root-owned and read-only, startup enforces it and refuses to start if the hash does not match.
- **Gateway token environment.** The gateway exports `OPENCLAW_GATEWAY_TOKEN` and writes it to `/tmp/nemoclaw-proxy-env.sh` for interactive sandbox sessions. Keep this in mind when deciding whether a workload should run with mutable config or an immutable config posture.

| Aspect | Detail |
|---|---|
| Default | The sandbox keeps `/sandbox/.openclaw` writable (`700 sandbox:sandbox`), sets `openclaw.json` to `600 sandbox:sandbox`, lets the agent manage state directly, and has the gateway place `OPENCLAW_GATEWAY_TOKEN` in `/tmp/nemoclaw-proxy-env.sh` for interactive shells. |
| Default | The sandbox keeps `/sandbox/.openclaw` writable (`2770 sandbox:sandbox`), sets `openclaw.json` to `660 sandbox:sandbox`, lets the agent manage state directly, and has the gateway place `OPENCLAW_GATEWAY_TOKEN` in `/tmp/nemoclaw-proxy-env.sh` for interactive shells. |
Comment on lines +185 to +191

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

⚠️ Potential issue | 🟠 Major | ⚡ Quick win

Do not edit autogenerated skill files directly.

This change is in an autogenerated skill file and should be produced by the docs-to-skills pipeline, not hand-edited in place.

As per coding guidelines: "User skills under .agents/skills/nemoclaw-user-*/*.md are autogenerated and must never be edited directly"

🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In @.agents/skills/nemoclaw-user-configure-security/references/best-practices.md
around lines 185 - 191, This file is autogenerated and must not be edited
directly; revert any manual edits to
.agents/skills/nemoclaw-user-configure-security/references/best-practices.md and
instead update the source documentation used by the docs-to-skills pipeline (the
upstream docs/skill generator) so the change is regenerated; ensure any
modifications to the DAC permissions, config-hash, or gateway token wording are
made in the canonical source for the docs-to-skills conversion and then run the
pipeline to produce the updated autogenerated markdown.

| What you can change | Apply a reviewed host-side immutability workflow to lock config and state directories with DAC permissions and the immutable flag where available. |
| Risk of default | A writable `.openclaw` directory lets the agent modify its own gateway config: disabling CORS or redirecting inference to an attacker-controlled endpoint. |
| Recommendation | For always-on assistants handling sensitive workloads, lock config after initial setup. For development workflows, the writable default is appropriate. |
Expand Down
1 change: 1 addition & 0 deletions .agents/skills/nemoclaw-user-manage-policy/SKILL.md
Original file line number Diff line number Diff line change
Expand Up @@ -175,6 +175,7 @@ Available presets:
| `jira` | Atlassian Jira API |
| `local-inference` | Local Ollama and vLLM through the host gateway |
| `npm` | npm and Yarn registries |
| `openclaw-pricing` | OpenClaw model-pricing reference fetch (LiteLLM and OpenRouter) |
| `outlook` | Microsoft 365 and Outlook |
| `pypi` | Python Package Index |
| `slack` | Slack API and webhooks |
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -45,6 +45,7 @@ NemoClaw ships maintained policy presets for common services in `nemoclaw-bluepr
| Hugging Face Hub and Inference API | `huggingface` |
| Jira and Atlassian Cloud | `jira` |
| Local Ollama or vLLM through the host gateway | `local-inference` |
| OpenClaw model-pricing reference fetch | `openclaw-pricing` |
| npm and Yarn packages | `npm` |
| Microsoft 365, Outlook, and Graph API | `outlook` |
| Python Package Index | `pypi` |
Expand Down Expand Up @@ -269,6 +270,23 @@ $ nemoclaw my-assistant exec -- brew install <formula>

You do not need to bootstrap Homebrew, install build dependencies, or source `brew shellenv` inside the sandbox.

## Model Pricing

OpenClaw's gateway fetches reference pricing from LiteLLM and OpenRouter on every start so it can populate `usage.cost` in session JSONL records.
The default-strict egress policy denies both hosts.
The fetch fails closed, the gateway logs `[gateway/model-pricing] LiteLLM pricing fetch failed: TypeError: fetch failed` (and the matching OpenRouter line) on every startup, and every session record records `usage.cost = 0` even though the input and output token counts populate correctly.
Tools that read the session log to display per-turn cost (audit dashboards, compliance review surfaces) cannot distinguish a real free run from this silent failure.

Apply the `openclaw-pricing` preset to allow both pricing endpoints.
The preset pins each host to a single read-only path so it does not widen egress beyond the pricing fetch:

```console
$ nemoclaw my-assistant policy-add openclaw-pricing --dry-run
$ nemoclaw my-assistant policy-add openclaw-pricing --yes
```

After the next gateway restart the WARN entries stop and `usage.cost` populates from the fetched pricing tables.

## Local Inference

Use `local-inference` when the sandbox needs access to host-side local inference services such as Ollama or vLLM through the OpenShell host gateway.
Expand Down
13 changes: 13 additions & 0 deletions .agents/skills/nemoclaw-user-overview/references/release-notes.md
Original file line number Diff line number Diff line change
Expand Up @@ -4,6 +4,19 @@

NVIDIA NemoClaw is available in early preview starting March 16, 2026. Use this page to track changes.

## v0.0.53

NemoClaw v0.0.53 focuses on safer sandbox recreation, stricter onboarding preflight defaults, local inference reliability, policy coverage, and day-two repair workflows:

- `nemoclaw onboard` backs up workspace state before deleting an existing sandbox during recreation, including sandboxes that are registered but not ready. If the backup is partial or fails, onboarding aborts before delete so workspace, skills, extensions, identity, memory, messaging state, and credentials are not silently dropped. Set `NEMOCLAW_RECREATE_WITHOUT_BACKUP=1` only when you intentionally want a fresh workspace.
- Under-provisioned container-runtime warnings now default to abort in interactive onboarding. Pressing Enter at the warning stops the run so you can resize Docker Desktop or Colima before the sandbox build stalls. Non-interactive runs continue with a warning, and `NEMOCLAW_IGNORE_RUNTIME_RESOURCES=1` still suppresses the check when you have already accepted the resource trade-off.
- OpenClaw sandboxes can use the new `openclaw-pricing` policy preset for model-pricing reference fetches from LiteLLM and OpenRouter. NemoClaw suggests this preset during OpenClaw onboarding so session JSONL records can populate `usage.cost` without widening egress beyond the two read-only pricing endpoints.
- Local Ollama onboarding is more accurate. NemoClaw validates the `/api/tags` response body through the authenticated proxy, honors accepted no-tools overrides through validation and proxy setup, and uses Ollama's reported runtime context length for `contextWindow` unless you set `NEMOCLAW_CONTEXT_WINDOW`.
- Onboarding and gateway reuse recover from more host-runtime drift. NemoClaw recovers stopped gateways before preserving PVC-backed state, verifies gateway containers before reusing port-conflict state, defers Docker-driver gateway teardown until step `[2/8]`, records Docker-driver sandboxes on macOS, and uses Docker `--gpus` rather than CDI repair on WSL Docker Desktop.
- The sandbox and integration paths handle more common failures cleanly, including Brave Search credential rewrite through OpenShell providers, Telegram placeholder repair, host-gateway `web_fetch` routing, read-only host targets for `share mount`, live gateway drift in `list`, host-alias Kubernetes invocations, Jetson bridge DNS preflight failures, and non-ready sandboxes during maintenance backups.
- Hermes startup no longer treats a fresh root-entrypoint layout as locked state, which avoids false locked-layout detection during sandbox boot.
- Maintainer tooling can export a signed skills catalog, detect untracked files during skills refresh diffs, and run the stale-issue verification workflow added for maintainers.

Comment on lines +7 to +19

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

⚠️ Potential issue | 🟠 Major | ⚡ Quick win

Do not edit autogenerated skill files directly.

This file is under the autogenerated skills path and should not be changed in PRs; keep docs edits in docs/ sources and let the docs-to-skills flow regenerate outputs.

As per coding guidelines: "User skills under .agents/skills/nemoclaw-user-*/*.md are autogenerated and must never be edited directly" and "For normal documentation changes, include only source pages under docs/."

🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In @.agents/skills/nemoclaw-user-overview/references/release-notes.md around
lines 7 - 19, The release-notes entry under "## v0.0.53" is in an autogenerated
skill file and must not be edited directly; revert any edits to that generated
file (the release-notes.md content) and instead apply your changes to the
canonical docs source (under docs/—e.g., the release notes or changelog source
page), then run the docs-to-skills generation/CI to regenerate
.agents/skills/nemoclaw-user-overview/references/release-notes.md so the update
appears in the generated artifact; ensure your PR removes direct changes to the
generated file and only includes changes to the source docs and the generated
output if produced by the regeneration step.

## v0.0.52

NemoClaw v0.0.52 upgrades the bundled OpenClaw runtime, repairs Hermes sandbox startup, restores onboarding ready output, and hardens Slack onboarding, Windows bootstrap, and private-network handling:
Expand Down
15 changes: 12 additions & 3 deletions .agents/skills/nemoclaw-user-reference/references/commands.md
Original file line number Diff line number Diff line change
Expand Up @@ -113,9 +113,8 @@ NemoClaw filters tier suggestions and resume selections by active agent support,
| `custom` | Apply exactly `NEMOCLAW_POLICY_PRESETS`. Previously-applied presets not in the list are removed. Alias: `list`. |
| `skip` | Skip the policy step entirely. Aliases: `none`, `no`. |

If you enable Brave Search during onboarding, NemoClaw currently stores the Brave API key in the sandbox's OpenClaw configuration.
That means the OpenClaw agent can read the key.
NemoClaw explores an OpenShell-hosted credential path first, but the current OpenClaw Brave runtime does not consume that path end to end yet.
If you enable Brave Search during onboarding, NemoClaw registers a Brave Search OpenShell provider and keeps `openclaw.json` on an OpenShell credential placeholder.

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

⚠️ Potential issue | 🟡 Minor | ⚡ Quick win

Fix phrasing for Brave credential placeholder.

Line 116 reads awkwardly: “keeps openclaw.json on an OpenShell credential placeholder.”
Use “as an OpenShell credential placeholder” (or equivalent) to avoid ambiguity.

🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In @.agents/skills/nemoclaw-user-reference/references/commands.md at line 116,
Update the phrasing that currently says “keeps `openclaw.json` on an OpenShell
credential placeholder” to use clearer wording such as “keeps `openclaw.json` as
an OpenShell credential placeholder”; locate the sentence containing the strings
`openclaw.json` and OpenShell in the commands.md content and replace “on an
OpenShell credential placeholder” with “as an OpenShell credential placeholder”
(or an equivalent construction) to remove ambiguity.

At egress, OpenShell rewrites Brave's `X-Subscription-Token` header with the real `BRAVE_API_KEY`.
Treat Brave Search as an explicit opt-in and use a dedicated low-privilege Brave key.

For non-interactive onboarding, you must explicitly accept the third-party software notice:
Expand Down Expand Up @@ -178,6 +177,13 @@ In interactive mode, the wizard asks for confirmation before delete and recreate
In non-interactive mode, NemoClaw recreates automatically when the stored selection is readable and differs; if NemoClaw cannot read the stored selection, NemoClaw reuses by default.
Set `NEMOCLAW_RECREATE_SANDBOX=1` to force recreation even when no drift is detected.

Before deleting an existing sandbox during recreation, NemoClaw backs up the workspace state (agents, extensions, workspace, skills, hooks, identity, devices, canvas, cron, memory, telegram, wechat, credentials) and restores it into the new sandbox once it is live.
This applies whether the existing sandbox is ready or marked not-ready, so cross-version upgrades that pass `NEMOCLAW_RECREATE_SANDBOX=1` no longer drop user files under `/sandbox/.openclaw/workspace/`.
The behaviour matches `nemoclaw <name> rebuild --force`.
NemoClaw aborts the recreate when the backup cannot complete in full — including when individual state directories or files fail mid-backup — so failed entries are not silently dropped on delete.
Set `NEMOCLAW_RECREATE_WITHOUT_BACKUP=1` to skip the pre-recreate backup.
The destination sandbox starts with a fresh workspace.

Before creating the gateway, the wizard runs preflight checks.
It verifies that Docker is reachable, warns on untested runtimes such as Podman, and prints host remediation guidance when prerequisites are missing.
The preflight also enforces the OpenShell version range declared in the blueprint (`min_openshell_version` and `max_openshell_version`).
Expand All @@ -187,6 +193,9 @@ If release metadata is unavailable, the installer uses its bundled fallback pin

When NemoClaw finds an existing gateway to reuse, it probes the host gateway HTTP endpoint before declaring the gateway reusable.
If the container is running but the upstream is still warming up (for example, immediately after a Docker daemon restart), NemoClaw rebuilds the gateway instead of trusting stale metadata.
On the Docker-driver gateway path, preflight stays read-only when it detects a stale gateway (for example, a Docker-driver runtime env hash drift).
It prints a `⚠ Gateway will be recreated when sandbox creation starts` notice and defers the actual teardown to step `[2/8] Starting OpenShell gateway`.
This means pressing `Ctrl+C` between preflight and step `[2/8]` leaves the running gateway and existing sandbox containers untouched, so `nemoclaw onboard` is safe to run just to check preflight output.
For Linux Docker-driver gateways, onboarding also checks that a helper container on the OpenShell Docker network can reach `host.openshell.internal:<gateway-port>`.
If a host firewall blocks that sandbox path, onboarding exits with a `sudo ufw allow from <subnet> to any port <gateway-port> proto tcp` command before it reports the gateway healthy.
Tune the wait via `NEMOCLAW_REUSE_HEALTH_POLL_COUNT` (default `6`) and `NEMOCLAW_REUSE_HEALTH_POLL_INTERVAL` (default `5` seconds).
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -397,6 +397,9 @@ Default Colima ships with 2 vCPU and 2 GiB of memory, which is not enough headro
On macOS Apple Silicon, the build can stall part-way through with no progress and no error, leaving the wizard waiting indefinitely.

Preflight inspects `docker info` for `NCPU` and `MemTotal` and prints a warning when the runtime falls below 4 vCPU or 8 GiB.
In interactive onboarding, the warning prompt defaults to abort, so pressing Enter stops the run before the sandbox build reaches the likely stall point.
Type `y` only when you intentionally want to continue on the smaller runtime.
Non-interactive onboarding prints the warning and continues.
On Colima, raise the resources before re-running onboard:
Comment on lines +400 to 403

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

⚠️ Potential issue | 🟠 Major | ⚡ Quick win

Generated skill content should come from source docs, not direct edits.

These updates are in an autogenerated skill file and should be produced by the docs-to-skills pipeline from docs/ changes.

As per coding guidelines: "User skills under .agents/skills/nemoclaw-user-*/*.md are autogenerated and must never be edited directly".

Also applies to: 780-795, 1190-1194, 1211-1215

🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In @.agents/skills/nemoclaw-user-reference/references/troubleshooting.md around
lines 400 - 403, Revert the manual edits in the autogenerated
nemoclaw-user-reference troubleshooting markdown (the skill file under the
agents skills set) and stop editing autogenerated skill files directly; instead
apply the content changes to the canonical source docs in the docs/ tree that
feed the docs-to-skills pipeline, then run the docs-to-skills pipeline to
regenerate the skill; also remove any direct edits in the other autogenerated
blocks referenced (the other occurrences around the same troubleshooting
content) so all updates originate from the source docs only.


```console
Expand Down Expand Up @@ -774,6 +777,23 @@ WhatsApp pairs entirely inside the sandbox.
NemoClaw advertises WhatsApp for OpenClaw and Hermes sandboxes after you add the channel on the host.
Run `openclaw channels login --channel whatsapp` inside OpenClaw sandboxes, or run `hermes whatsapp` inside Hermes sandboxes.

### `scripts/rcf_patch.py` is missing from the blueprint

`scripts/rcf_patch.py` is intentionally absent from current NemoClaw blueprints.
Older QA plans used that helper for a Dockerfile "Patch-4" test that corrupted the build-time `replaceConfigFile` monkey-patch and expected `ERROR: Patch 4 (replaceConfigFile EACCES) not applied`.
The old Patch-4 fail-closed test no longer applies because NemoClaw no longer patches OpenClaw's compiled `replaceConfigFile` source at image build time.

Current sandboxes use a mutable-default config model instead.
Before a reviewed host-side lockdown, `/sandbox/.openclaw/openclaw.json` is group-writable by the sandbox and gateway users, so OpenClaw config mutations should write normally rather than requiring an EACCES swallow.
After lockdown, runtime config mutations should fail cleanly or route users to the supported host-side NemoClaw command.

To validate this area now, use the config lifecycle tests instead of looking for `rcf_patch.py`:

```console
$ npm run build:cli
$ npm test -- test/repro-2681-group-writable.test.ts
```

### `openclaw config set` or `unset` is blocked inside the sandbox

This is expected.
Expand Down Expand Up @@ -1167,7 +1187,10 @@ Recent NVIDIA Container Toolkit installs configure the Docker daemon for Contain
If no `nvidia.com/gpu` CDI spec has been generated on the host yet, gateway start fails with `Docker responded with status code 500: CDI device injection failed: unresolvable CDI devices nvidia.com/gpu=all`.
The standard NemoClaw installer detects this gap before onboarding, first tries to enable the NVIDIA CDI refresh systemd units, and falls back to generating the spec directly with `nvidia-ctk`.
If you run `nemoclaw onboard` directly, preflight prints the manual remediation instead.
The underlying fix is the same on any Docker host whose `docker info` advertises a non-empty `CDISpecDirs`.
The native Linux fix is the same on Docker hosts whose `docker info` advertises a non-empty `CDISpecDirs`.
On WSL with Docker Desktop, Docker may advertise CDI directories even though `--device nvidia.com/gpu=all` is not usable from the WSL distro.
For that runtime, NemoClaw skips Linux CDI repair and uses Docker's `--gpus` compatibility path for sandbox GPU access.
This compatibility path can be retired once Docker Desktop exposes usable `nvidia.com/gpu` CDI specs inside WSL, or once OpenShell no longer requires host-visible CDI specs for Docker Desktop WSL GPU passthrough.

Enable the refresh units, verify they list `nvidia.com/gpu` entries, then rerun onboarding:

Expand All @@ -1185,6 +1208,12 @@ $ sudo nvidia-ctk cdi generate --output=/etc/cdi/nvidia.yaml
$ nvidia-ctk cdi list
```

On WSL with Docker Desktop, confirm Docker Desktop WSL integration is enabled for your distro and verify Docker GPU access from WSL:

```console
$ docker run --rm --gpus all nvcr.io/nvidia/k8s/cuda-sample:nbody nbody -gpu -benchmark
```

If GPU passthrough is not required on this host, rerun onboarding with `--no-gpu` instead.

### Docker GPU patch failed during sandbox create
Expand Down
Loading
Loading