fix(connect): show agent-specific command and /exit hint (Fixes #2079)

[kagura-agent]

Summary

Fixes #2079 — the connect hint showed incorrect instructions after connecting to a sandbox.

Problems

1. Wrong TUI command: Always showed openclaw tui regardless of which agent the sandbox runs. For hermes sandboxes, the correct command is hermes.
2. Missing /exit hint: The exit hint only mentioned exit (shell exit), but users who launched the TUI first need /exit to leave the chat before exit returns them to the host shell.

Changes

• src/nemoclaw.ts: Derive the TUI command from the sandbox registry's agent field (hermes → hermes, others → openclaw tui). Updated exit hint to mention /exit for leaving the chat.
• scripts/install.sh: Show agent-specific command in post-onboard hint (was only shown for openclaw agent, now shown for all agents with correct command).

Testing

• Ran full test suite — 0 new failures (38 pre-existing failures unchanged)
• No test assertions reference the hint text (tests suppress via NEMOCLAW_NO_CONNECT_HINT=1)

Signed-off-by: kagura-agent [email protected]

Summary by CodeRabbit

• Bug Fixes

  • Sandbox connection hint now always shows a single connect command reflecting the configured agent (Hermes, OpenClaw tui, or the agent's name).

• Documentation

  • Clarified exit instructions: use "/exit" to leave the in-sandbox chat, then "exit" to return to the host shell.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro Plus

Run ID: 12647975-c7c0-4558-ac63-c4ea6b0d3fe1

📥 Commits

Reviewing files that changed from the base of the PR and between 577c1ea and bf64357.

📒 Files selected for processing (2)

• scripts/install.sh
• src/nemoclaw.ts

🚧 Files skipped from review as they are similar to previous changes (2)

• scripts/install.sh
• src/nemoclaw.ts

---

📝 Walkthrough

Walkthrough

Post-connection hints were made agent-aware: the suggested in-sandbox command now reflects the sandbox's configured agent (e.g., hermes or openclaw tui) and the exit guidance was changed to instruct /exit to leave the chat, then exit to return to the host shell.

Changes

Cohort / File(s)	Summary
Agent-specific connection & exit hints scripts/install.sh, src/nemoclaw.ts	Compute and print the in-sandbox agent command from the sandbox agent setting (prints hermes for hermes, openclaw tui for openclaw, or the raw agent name otherwise). Update exit guidance to: "Type /exit to leave the chat, then exit to return to the host shell."

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~8 minutes

Poem

A rabbit peeks in, polite and spry,
"Run hermes now," it whispers sly.
When chat is done, don't quit in haste—
/exit then exit, no trace misplaced. 🐰✨

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 50.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly and concisely summarizes the main changes: fixing the connect hint to show the correct agent-specific command and the /exit instruction.
Linked Issues check	✅ Passed	The code changes directly address both objectives from issue #2079: showing agent-specific commands (hermes instead of openclaw tui) and clarifying the /exit hint for leaving the chat.
Out of Scope Changes check	✅ Passed	All changes are directly related to the objectives in issue #2079; no out-of-scope modifications were introduced.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[coderabbitai]

Actionable comments posted: 2

🤖 Prompt for all review comments with AI agents

Verify each finding against the current code and only fix it if needed.

Inline comments:
In `@scripts/install.sh`:
- Around line 263-267: The printf branch currently only handles agent_name ==
"hermes" and else prints "openclaw tui" for all other agents, which gives
incorrect post-install command hints; update the conditional logic in the
install.sh section handling agent_name (the if checking "hermes" and the else
printf lines) to cover other known agents or default to a generic command hint:
add explicit branches for additional agent names you support (e.g., "openclaw",
"otherAgentName") each printing the correct sandbox command, and change the
fallback printf to a generic message that includes the actual agent_name
variable (e.g., "sandbox@%s$ %s") so unknown values show the agent_name rather
than always "openclaw tui".

In `@src/nemoclaw.ts`:
- Around line 1233-1238: The hint hardcodes "openclaw tui" for all non-Hermes
agents; change agentCmd construction to derive the CLI from sb?.agent so it
correctly reflects other agent names. Specifically, update the agentCmd logic
(variable name agentCmd and the expression using sb?.agent) to: use "hermes"
when sb?.agent === "hermes", otherwise if sb?.agent is present use `${sb.agent}
tui` (so the hint shows e.g. "otherAgent tui"), and finally fallback to
"openclaw tui" only when sb?.agent is missing/undefined; ensure the rest of the
console.log lines continue to use agentCmd.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro Plus

Run ID: af1988ad-4f3c-49eb-b710-0dad5a88ad0c

📥 Commits

Reviewing files that changed from the base of the PR and between 2a29b5e and 577c1ea.

📒 Files selected for processing (2)

• scripts/install.sh
• src/nemoclaw.ts

[wscurran]

✨ Thanks for submitting this pull request that proposes a fix to the connect hint and adds a new exit hint, which could help improve the user experience of NemoClaw.

---

Possibly related open issues:

• #2079 Wrong instructions after connection established to hermes sandboxes

[ericksoa]

Reviewed: small, well-scoped UX fix. Logic is consistent between TS and shell implementations, properly derives agent-specific commands, and improves exit instructions. LGTM.