Improve external agent invite flow (#6183)

## Thinking Path

> - Paperclip orchestrates AI agents for zero-human companies.
> - Agent creation can happen through local runtimes, managed runtimes,
and external agents that onboard through invites.
> - The old OpenClaw-oriented invite UX lived under company
settings/invites and made a gateway-specific path look like a company
access setting.
> - That hid the broader bring-your-own-agent flow and forced operators
to leave the add-agent modal when adding an external agent.
> - This pull request moves external agent invite generation into the
add-agent modal and makes the copy agent-oriented instead of
OpenClaw-only.
> - The benefit is a clearer agent-first onboarding path while company
invites stay focused on human access.

## What Changed

- Added an external-agent invite branch to the add-agent modal,
including a dedicated prompt result view with Back navigation.
- Added a shared agent onboarding prompt builder and focused modal
coverage for prompt replacement/back navigation.
- Removed the agent invite prompt UI from Company Settings and Company
Invites, leaving Company Invites focused on human access links and
invite history.
- Updated the hidden OpenClaw Gateway runtime hint to direct operators
to the add-agent invite flow instead of presenting it as a blocked
runtime card.
- Updated invite/onboarding docs, storybook coverage, and server-side
onboarding copy toward generic agent language while preserving existing
gateway compatibility.

## Verification

- `pnpm -r typecheck`
- `pnpm build`
- `FAKE_BIN="$(mktemp -d)/bin"; mkdir -p "$FAKE_BIN"; printf
'#!/bin/sh\nexit 1\n' > "$FAKE_BIN/tailscale"; chmod +x
"$FAKE_BIN/tailscale"; PATH="$FAKE_BIN:$PATH" pnpm test:run`
- `pnpm test:run` without the fake `tailscale` shim was also attempted;
it failed only in two pre-existing CLI tailnet fallback tests because
this host has a real Tailscale address (`100.125.202.3`) where those
tests expect no Tailscale.
- Focused confirmation for that host-env issue: `FAKE_BIN=...
PATH="$FAKE_BIN:$PATH" pnpm exec vitest run --project paperclipai
cli/src/__tests__/network-bind.test.ts
cli/src/__tests__/onboard.test.ts`
- Manual UI verification: served UI locally in light mode, opened
add-agent modal, generated external agent prompt, verified the generated
prompt replaces the form and Back returns to the form.

### Screenshots

![Add agent
modal](https://raw.githubusercontent.com/aronprins/paperclip/pr-assets/6183-agent-invites/.github/pr-screenshots/6183/add-agent-modal-light.png)

![External agent invite
form](https://raw.githubusercontent.com/aronprins/paperclip/pr-assets/6183-agent-invites/.github/pr-screenshots/6183/external-agent-invite-form-light.png)

![Generated onboarding prompt replacement
view](https://raw.githubusercontent.com/aronprins/paperclip/pr-assets/6183-agent-invites/.github/pr-screenshots/6183/onboarding-prompt-result-light.png)

## Risks

- Existing OpenClaw gateway compatibility remains, but operators now
discover external agent onboarding from the add-agent modal instead of
company settings.
- Agent invites still appear in the invite history table, so that page
may show agent-scoped invite rows even though it no longer creates agent
onboarding prompts.
- Low migration risk: no schema changes.

> For core feature work, check [`ROADMAP.md`](ROADMAP.md) first and
discuss it in `#dev` before opening the PR. Feature PRs that overlap
with planned core work may need to be redirected — check the roadmap
first. See `CONTRIBUTING.md`.

## Model Used

- OpenAI Codex, GPT-5 coding agent in Codex desktop; tool-enabled
repository, shell, browser, and GitHub workflow. Context window size was
not exposed by the runtime.

## Checklist

- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have run tests locally and they pass
- [x] I have added or updated tests where applicable
- [x] If this change affects the UI, I have included before/after
screenshots
- [x] I have updated relevant documentation to reflect my changes
- [x] I have considered and documented any risks above
- [x] I will address all Greptile and reviewer comments before
requesting merge

doc/DEVELOPING.md

@@ -554,10 +554,12 @@ pnpm paperclipai dashboard get
 
 See full command reference in `doc/CLI.md`.
 
-## OpenClaw Invite Onboarding Endpoints
+## Agent Invite Onboarding Endpoints
 
 Agent-oriented invite onboarding now exposes machine-readable API docs:
 
+The board UI generates agent onboarding prompts from the add-agent modal (`+` in the agent sidebar), so agent onboarding sits with the rest of agent creation rather than company member invite settings.
+
 - `GET /api/invites/:token` returns invite summary plus onboarding and skills index links.
 - `GET /api/invites/:token/onboarding` returns onboarding manifest details (registration endpoint, claim endpoint template, skill install hints).
 - `GET /api/invites/:token/onboarding.txt` returns a plain-text onboarding doc intended for both human operators and agents (llm.txt-style handoff), including optional inviter message and suggested network host candidates.
@@ -575,7 +577,7 @@ pnpm smoke:openclaw-join
 What it validates:
 
 - invite creation for agent-only join
-- agent join request using `adapterType=openclaw`
+- agent join request using `adapterType=openclaw_gateway`
 - board approval + one-time API key claim semantics
 - callback delivery on wakeup to a dockerized OpenClaw-style webhook receiver