## Summary

Bumps the published doc version to `0.0.22` and documents the
user-visible CLI behavior changes to `nemoclaw <name> connect` that
landed since v0.0.21. Drafted via the `nemoclaw-contributor-update-docs`
skill against commits in `v0.0.21..origin/main`, filtered through
`docs/.docs-skip`.

## Changes

- **`docs/project.json`** and **`docs/versions1.json`**: bump the
published version from `0.0.20` to `0.0.22`; insert a `0.0.21` entry
into the version list so the history stays contiguous.
- **`docs/reference/commands.md`** → `nemoclaw <name> connect`: document
two new behaviors.
- Readiness poll with `NEMOCLAW_CONNECT_TIMEOUT` (integer seconds;
default `120`) that replaces the silent hang when the sandbox is not yet
`Ready` — right after onboarding, while the 2.4 GB image is still
pulling (#466).
- Post-connect hint is now agent-aware, names the correct TUI command
for the sandbox's agent, and tells you to use `/exit` to leave the chat
before `exit` returns you to the host shell (#2080).

Feature PRs that shipped their own docs in the same commit are
intentionally not re-documented here:

- `channels list/add/remove` (#2139) — command reference and the
"`openclaw channels` blocked inside the sandbox" troubleshooting entry
landed with the feature.
- `nemoclaw gc` (#2176) — documented as part of the destroy/rebuild
image cleanup PR.

Skipped per `docs/.docs-skip`:

- `e6bad533 fix(shields): verify config lock and fail hard on re-lock
failure (#2066)` — matched `skip-features: src/lib/shields.ts`.

Other commits in the range (#2141 OpenShell version bump, #1819 plugin
banner live inference probe, #2085 / #2146 Slack Socket Mode fixes,
#2110 axios proxy fix, #1818 NIM curl timeouts, #1824 onboard gateway
bootstrap recovery, and assorted CI / test / install plumbing) are
internal behavior refinements with no doc-relevant surface change.

## Type of Change

- [ ] Code change (feature, bug fix, or refactor)
- [ ] Code change with doc updates
- [ ] Doc only (prose changes, no code sample modifications)
- [x] Doc only (includes code sample changes)

## Verification

- [x] `npx prek run --all-files` passes for the modified files via the
pre-commit hook, including `Regenerate agent skills from docs` (source ↔
generated parity confirmed)
- [ ] `npm test` passes — skipped; the one pre-existing
`test/cli.test.ts > unknown command exits 1` failure on `origin/main` is
unrelated to these markdown/JSON-only changes
- [ ] Tests added or updated for new or changed behavior — n/a, doc-only
- [x] No secrets, API keys, or credentials committed
- [x] Docs updated for user-facing behavior changes
- [ ] `make docs` builds without warnings (doc changes only) — not run
locally
- [x] Doc pages follow the [style
guide](https://github.com/NVIDIA/NemoClaw/blob/main/docs/CONTRIBUTING.md)
(doc changes only)
- [ ] New doc pages include SPDX header and frontmatter (new pages only)
— n/a, no new pages

## AI Disclosure

- [x] AI-assisted — tool: Claude Code

---
Signed-off-by: Miyoung Choi <miyoungc@nvidia.com>

<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit

* **New Features**
* `connect` now displays the sandbox phase while waiting for readiness
and honors a configurable timeout via NEMOCLAW_CONNECT_TIMEOUT (default
120s).
* TTY hints are agent-aware and instruct using `/exit` before returning
to the host shell.

* **Documentation**
  * Command docs updated to describe polling, timeout, and TTY guidance.
* Project/docs metadata updated for versions 0.0.21 and 0.0.22 (package
version bumped to 0.0.22).
<!-- end of auto-generated comment: release notes by coderabbit.ai -->

---------

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

#2592) (#3009)

## Summary

#2592 reports that `openclaw channels add telegram` and `openclaw
channels remove telegram` hit OpenClaw's CLI parser instead of the
in-sandbox guard added in #2139. The guard logic in
`scripts/nemoclaw-start.sh` is generic — it treats `add/remove
<anything>` as mutating — so the implementation already covers the
reporter's exact scenarios. What's missing is per-channel test coverage:
the existing configure-guard test only exercises `openclaw channels add
slack`, leaving the telegram/discord rows and the entire `remove` op
unverified.

## Problem

The configure-guard test in `test/nemoclaw-start.test.ts` currently
asserts one combo. If a future refactor of the `case "$2" in` block in
`scripts/nemoclaw-start.sh` accidentally narrows the match (e.g., adding
an explicit allowlist that misses telegram, or restricting the inner
case to specific verbs), only the `add slack` path is locked in. The
reporter's complaint is exactly that shape — they hit a path that wasn't
covered by the existing test.

## Changes

- Add a parametric test under the same `nemoclaw-start configure guard
behavior` describe block that asserts the guard fires for every `(slack,
telegram, discord) x (add, remove)` combination — six cases total.
- Each case asserts the guard returned non-zero status, surfaced
`openclaw channels <op>` in stderr, and pointed at the host-side hint
`nemoclaw <sandbox> channels <op>`.

No `scripts/nemoclaw-start.sh` changes — the guard at
`scripts/nemoclaw-start.sh:2101-2117` already covers all six combos.

## Type of Change

- [x] Code change (feature, bug fix, or refactor)

## Test plan

- [x] `npx vitest run test/nemoclaw-start.test.ts -t "#2592"` — 1/1 pass
- [x] No secrets, API keys, or credentials committed
- [x] Tests added for new behavior

---
Signed-off-by: latenighthackathon
<latenighthackathon@users.noreply.github.com>

<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->

## Summary by CodeRabbit

* **Tests**
* Expanded test coverage for configuration guard behavior validation
across multiple scenarios.

<!-- end of auto-generated comment: release notes by coderabbit.ai -->

Signed-off-by: latenighthackathon <latenighthackathon@users.noreply.github.com>
Co-authored-by: latenighthackathon <latenighthackathon@users.noreply.github.com>