@sylphx/flow 2.7.0 → 2.9.0

package/CHANGELOG.md

@@ -1,5 +1,24 @@
 # @sylphx/flow
 
+## 2.9.0 (2025-12-17)
+
+Add comprehensive SaaS review command suite with parallel worker delegation.
+
+- `/saas-admin`: Admin platform review (RBAC, bootstrap, config, feature flags, ops)
+- Updated `/saas-review`: All 8 workers run in parallel (domains + discovery)
+- Clarified Task tool delegation pattern for worker spawning
+
+### 📚 Documentation
+
+- **commands:** run all 8 workers in parallel (merge discovery into Phase 1) ([cd0afd6](https://github.com/SylphxAI/flow/commit/cd0afd630d6c0d4f751cec81cce4572b19bcbe9d))
+- **commands:** clarify Task tool delegation pattern in saas-review ([2940cb0](https://github.com/SylphxAI/flow/commit/2940cb094fb3ce65b0af59f3c1541769576c1050))
+
+## 2.8.0 (2025-12-17)
+
+### ✨ Features
+
+- **commands:** add /saas-admin for admin platform review ([f67b74f](https://github.com/SylphxAI/flow/commit/f67b74f2cac978274ad546cc7c70912f3cd4700c))
+
 ## 2.7.0 (2025-12-17)
 
 ### ✨ Features

package/assets/slash-commands/saas-admin.md

@@ -0,0 +1,123 @@
+---
+name: saas-admin
+description: SaaS admin platform review - RBAC, bootstrap, config, feature flags, ops
+agent: coder
+---
+
+# Admin Platform Review
+
+## Scope
+
+Review admin systems: RBAC, bootstrap flow, configuration management, feature flags governance, operational tooling, and impersonation.
+
+## Specification
+
+### Access Control (RBAC)
+
+* **Least privilege principle**: Users get minimum permissions needed.
+* Role hierarchy with clear inheritance.
+* Permission granularity (resource-level, action-level).
+* All authorization is **server-enforced**; no client-trust.
+* Role changes require appropriate privilege level and are audited.
+
+### Admin Bootstrap (Hard Requirement)
+
+* Admin bootstrap must **not rely on file seeding**.
+* Use a secure, auditable **first-login allowlist** for the initial SUPER_ADMIN.
+* **Permanently disable bootstrap** after completion — no re-entry.
+* All privilege grants must be server-enforced and recorded in the audit log.
+* The allowlist must be managed via **secure configuration (environment/secret store)**, not code or DB seeding.
+
+### Configuration Management
+
+* All **non-secret** product-level configuration must be manageable via admin (server-enforced).
+* Configuration changes require **validation and change history**.
+* Secrets/credentials are **environment-managed only**; admin may expose safe readiness/health visibility, not raw secrets.
+* Support for environment-specific overrides (dev/staging/prod).
+* Rollback capability for configuration changes.
+
+### Feature Flags Governance
+
+* Gradual rollout support (percentage-based, user segment-based).
+* A/B testing integration where applicable.
+* **Audit trail** for all flag changes (who/when/why).
+* Emergency **kill switches** for rapid disable.
+* Flag lifecycle management (created → active → deprecated → removed).
+* Server-enforced evaluation; no client-side flag source-of-truth.
+
+### Operational Management
+
+* **User/account management tools**:
+  * Search, view, edit user profiles
+  * Account status management (active, suspended, banned)
+  * Manual verification/unverification
+
+* **Entitlements/access management**:
+  * View and modify user entitlements
+  * Grant/revoke access with audit trail
+  * Bulk operations with safeguards
+
+* **Lifecycle actions**:
+  * Account suspension/reactivation
+  * Data export (for user requests)
+  * Account deletion with proper cascade
+
+* **Issue resolution workflows**:
+  * Support ticket integration
+  * Action history per user
+  * Notes and annotations
+
+* **Step-up controls** for sensitive actions:
+  * Actions affecting money/credits require MFA
+  * Actions affecting security posture require MFA
+  * Destructive actions require confirmation + reason
+
+### Impersonation
+
+* Impersonation allowed **with explicit safeguards**:
+  * Requires elevated privilege level
+  * Time-limited sessions (auto-expire)
+  * Full audit logging (start, actions, end)
+  * Clear indicator in UI during impersonation
+  * Cannot impersonate higher-privilege users
+* All actions during impersonation attributed to both impersonator and target.
+* Optional: Visible indicator to impersonated user that session was accessed.
+
+### Admin Audit Logging
+
+* **All admin actions must be auditable**:
+  * Who performed the action
+  * When (timestamp with timezone)
+  * What action was taken
+  * Why (required reason for sensitive actions)
+  * Before/after state for mutations
+  * Correlation to session/request
+* Audit logs must be:
+  * Immutable (append-only)
+  * Queryable and filterable
+  * Exportable for compliance
+  * Retained per data retention policy
+
+## Domain Discovery
+
+After reviewing compliance with spec, explore improvements:
+
+* **Admin UX**: Is the admin panel efficient for common tasks? Keyboard shortcuts? Bulk actions?
+* **Self-service vs admin**: What admin actions could be self-service for users?
+* **Automation**: What repetitive admin tasks could be automated? Scheduled jobs?
+* **Alerting**: Should certain admin actions trigger alerts? (e.g., mass deletions)
+* **Delegation**: Can some admin tasks be delegated to lower roles safely?
+* **Mobile admin**: Is there a need for mobile admin access? How to secure?
+
+## Domain Gates
+
+* [ ] RBAC implemented with least privilege
+* [ ] Bootstrap flow is secure and one-time only
+* [ ] Config changes are validated and audited
+* [ ] Feature flags have full audit trail
+* [ ] Sensitive actions require step-up (MFA)
+* [ ] Impersonation is time-limited and fully logged
+* [ ] All admin actions are auditable
+* [ ] Audit logs are immutable and queryable
+* [ ] No hardcoded admin credentials anywhere
+* [ ] Admin endpoints are rate-limited

package/assets/slash-commands/saas-review.md

@@ -45,9 +45,16 @@ agent: coder
 
 ## Review Execution
 
-### Phase 1: Domain Reviews (Parallel)
+### Phase 1: All Reviews (Parallel)
 
-Delegate each domain to a worker agent. Workers should review compliance with spec AND identify domain-specific improvement opportunities.
+Spawn **all workers in parallel** using the Task tool. Each worker runs its slash command and returns findings.
+
+**Delegation pattern:**
+```
+Use Task tool with subagent_type: "Coder" for each worker.
+Spawn ALL 8 workers in a single message (parallel execution).
+Each worker prompt: "Run /{command} and return findings."
+```
 
 | Worker | Command | Focus |
 |--------|---------|-------|
@@ -57,16 +64,10 @@ Delegate each domain to a worker agent. Workers should review compliance with sp
 | Platform | `/saas-platform` | Design system, SEO, PWA, performance, a11y |
 | Security | `/saas-security` | OWASP, privacy, consent, observability, operability |
 | Growth | `/saas-growth` | Onboarding, referral, retention, guidance |
-
-### Phase 2: Strategic Discovery
-
-After domain reviews complete, run cross-domain strategic analysis:
-
-| Worker | Command | Focus |
-|--------|---------|-------|
+| Admin | `/saas-admin` | RBAC, bootstrap, config, feature flags, ops tooling |
 | Discovery | `/saas-discovery` | Feature opportunities, pricing optimization, competitive research |
 
-### Phase 3: Final Gate (You)
+### Phase 2: Final Gate (You)
 
 Synthesize all domain findings and discovery insights:
 
@@ -170,8 +171,7 @@ Checklist with pass/fail for each gate
 ## Completion Criteria
 
 Complete only when:
-- [ ] All domain reviews finished
-- [ ] Discovery phase completed
+- [ ] All 8 workers finished (domains + discovery)
 - [ ] All findings synthesized
 - [ ] Delivery gates verified
 - [ ] Integrated report produced

package/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@sylphx/flow",
-	"version": "2.7.0",
+	"version": "2.9.0",
 	"description": "One CLI to rule them all. Unified orchestration layer for Claude Code, OpenCode, Cursor and all AI development tools. Auto-detection, auto-installation, auto-upgrade.",
 	"type": "module",
 	"bin": {