feat: add public Agent Template library and onboarding intent

[yuezengwu]

Summary

• Add public /templates Library and /templates/:slug detail routes over the existing public-safe catalog projection, including active, retired/replacement, empty, missing, loading, and recoverable error states.
• Preserve the byte-exact canonical use intent /templates/:slug?use=1 through GitHub/Google login, while rejecting extra, duplicate, encoded, absolute, hash, backslash, trailing-separator, and trailing-slash variants.
• Resolve authenticated intent against an explicitly chosen Team. A switch proceeds only after a fresh authoritative /me snapshot confirms the exact membership; transport failure rolls back to the last confirmed Team, final 401/logout cannot resurrect old org state, and persistent ChatById failures do not create automatic retry loops.
• Carry a per-Team Template handoff through onboarding or preselect the existing New Agent dialog for established accounts. Creation passes the submit-time Template id through the existing atomic Agent POST; stale/retired/fetch-failed intents degrade to plain creation without submitting stale ids.
• Add WCAG-AA semantic foreground tokens for brand-filled actions and instrument Library/detail/use-started/create-success events without exposing private Template content.

Boundaries

• Builds on the Agent Template lifecycle shipped in feat: add Agent Template lifecycle surfaces #2130. Template use remains a one-time import into the selected Team; imported Team Resources and Agent bindings are the execution authority afterward.
• Public responses render only safe purpose, audience, value, and capability summaries. Prompt source, Skill attachment identity/content, MCP connection details, drafts, and malformed definitions stay private or fail closed.
• No database migration, Client materialization change, production catalog content, hidden prompt, automatic chat/message, second adoption request, catalog auto-sync, or activation claim.
• Agent creation still opens /?c=draft&with=<uuid>; a real first work loop remains a separate post-merge staging pilot.

Validation

• pnpm check — passed with 16 baseline warnings and no errors.
• pnpm typecheck — forced no-cache, 11/11 tasks.
• pnpm test — forced single-fork full monorepo run, 12/12 tasks; 912 files passed / 1 skipped, 10,105 tests passed / 4 skipped.
• Production build — forced no-cache, 5/5 tasks.
• Independent exact-head QA on 8270c4ac47bb63aee5592853feb3aa979dfb3d07 using an isolated production Server/Web/PostgreSQL/browser cell.

Independent QA exercised anonymous discovery and leakage checks, strict login intent, exact-Team selection, deterministic /me abort/retry and final-401 logout, ChatById retry bounds, per-org onboarding handoff, real Template and zero-Template Agent POSTs with API/DB readback, stale/failure recovery, and WCAG 2 AA across the new public controls. Brand CTA contrast measured 8.25:1 normal/focus and 5.58:1 hover.

Change Surface

• apps/cli public CLI or help output
• tree onboarding / binding / inspection behavior
• shipped or planned skill topology
• docs or contributor-facing repository metadata
• CI / packaging / release plumbing

Notes

• Package or install behavior changes: none.
• Docs or tests updated to match: focused Shared/Server/Web contracts plus real-browser QA; no committed QA case changed in this PR.
• Limitations: real third-party GitHub/Google account callback completion was unavailable; exact local OAuth start/state/next boundaries were validated. Creation and draft navigation intentionally do not simulate or claim organic runtime activation.
• Follow-up work: after merge, run the final staging pilot through anonymous discovery → auth/onboarding → creation → a real first task → Agent Detail/removal, plus ordinary-creation regression, before adding production catalog content.

[github-actions]

Thank you for your contribution. Before it can be accepted, you must be covered by the First Tree ICLA v1.0. Sign individually with the exact comment below. If you contribute for an organization, an authorized signatory can instead complete the First Tree CCLA v1.0 and send it to legal@first-tree.ai for verification.

---

I have read the CLA Document and I hereby sign the CLA

---

kiven-fte seems not to be a GitHub user. You need a GitHub account to be able to sign the CLA. If you have already a GitHub account, please add the email address used for this commit to your account.
_{You can retrigger this bot by commenting recheck in this Pull Request. Posted by the CLA Assistant Lite bot.}

[baixiaohang]

Recommendation: request changes

• Rationale: Two asynchronous state races can either discard an explicit Template adoption or let a stale /me response overwrite the current session.

Risk level: B-low

• Path baseline: includes packages/server/**, but the touched service is oauth-bootstrap rather than a B-high core service.
• Semantic lift: the Web auth/data layer and cross-Team selection logic independently lift the Web portion to B-low; no A/B-high path is touched.

PR summary

• Author / repo: yuezengwu / agent-team-foundation/first-tree
• Problem: Let anonymous visitors discover official Agent Templates, preserve a chosen Template through sign-in, and create a Template-backed Agent in an explicitly confirmed Team or through that Team's onboarding flow.
• Approach: Adds public Library/detail routes, a strict canonical OAuth intent, authoritative Team confirmation, per-Team onboarding handoff, Template preselection in both creation surfaces, safe analytics, and AA foreground tokens.
• Impacted modules: packages/shared, Server OAuth bootstrap, Web auth/Team selection, public Template pages, onboarding, New Agent creation, workspace chat routing, and design tokens.

Review findings
❌ 1. An explicit “Use this template” intent can silently become a plain Agent while its Template lookup is still pending. NewAgentDialog enables submit without waiting for templateCatalogLoaded, and StepCreateAgent enables create while intentSlug exists but intentTemplate has not resolved. In either surface, a slow/hung request lets the user submit templateIds: []; onboarding then clears the handoff after creation, so the requested responsibility is permanently lost without the user removing it or seeing an unavailable state. Keep creation disabled (with a loading state) while an explicit intent is unresolved, and permit plain creation only after a definitive unavailable/failure result or an explicit Remove. [R4 / packages/web/src/components/new-agent-dialog.tsx:741, packages/web/src/pages/onboarding/steps/step-create-agent.tsx:123]

❌ 2. Successful /me responses are not guarded against a logout or identity change that occurs while the request is in flight. selectOrganization records sessionMarker, but checks it only in the rejection path; a request started as user A can still return successfully after logout or after user B adopts tokens, and loadMe will repopulate user, memberships, confirmedOrgIdRef, the persisted selection, and the API org target from A. The new 401 test cannot cover this because it exercises only rejection. Capture the session/generation at request start and discard a successful response when it no longer matches before applying any state. [R4 / packages/web/src/auth/auth-context.tsx:293]

✅ The strict shared intent parser, public-safe Template DTO boundary, exact-Team membership check, and no-auto-retry ChatById guard are otherwise aligned with the documented cross-surface constraints.

Action taken

• Submitted request changes.

[yuezengwu]

Superseded by the CLA-valid replacement PR: #2134

The repaired source from this PR was transferred onto current main as one yuezengwu-authored commit, without amending or force-pushing this branch. The aggregate patch-id remains exactly 4f39a40b0bc7bbb3bee3be25ea6c11e063419b05; old repaired head c881569b32d13739f7f88a17a66da1aba6303d6b and replacement head b0744b78232f80efdaab2a5d41ef3f0fe91faae7 are source-equivalent across the candidate paths.

The replacement includes both requested review fixes (fail-closed unresolved Template intent and stale successful /me session isolation) plus their regression coverage. Closing this PR so CI, review, CLA, and fresh exact-head QA continue on the replacement only.