beifeng724/personal-workbench-skills / skills/personal-workbench-builder
personal-workbench-builder
VerifiedBuild a durable personal or professional workbench as a runnable application, or produce a six-stage generic prompt pack for building it. Use when the user wants to turn a recurring workflow into a dashboard, management system, workspace, personal operating system, or long-lived website with multiple connected modules and persistent state. Trigger for project, CRM, creator, learning, family, finance, travel, teaching, legal, or team workbenches. Do not trigger for marketing sites, one-off scripts, ordinary coding tasks, or informational questions.
references/prompt-pack.md
View on GitHub# Six-Stage Generic Prompt Pack
Generate prompts that work across capable coding agents. Do not name or optimize for WorkBuddy, Codex, Claude, Cursor, or another target tool.
## Shared context envelope
Put this at the beginning of every stage:
- Attach or paste the current `workbench-spec.md`.
- State the stage name and the previous stage's verified outputs.
- Require the agent to preserve all `confirmed` requirements.
- Require proposals, conflicts, and deferrals to be labeled; no silent scope changes.
- Require actual test evidence and explicit limitations.
- Require the current workflow state and `workflow_guard.py` result before the next state.
- Carry devices, synchronization, data importance, persistence, deployment, version, and recovery decisions forward.
- For L2-L4, require confirmation before material scope changes.
- For L4, prohibit implementation until the confirmation log contains explicit complete approval.
## Stage 1: Requirements
Instruct the agent to fill gaps progressively, classify L1-L4, run privacy review, create the requirement ledger, and obtain the required approval. Output the updated spec, unanswered questions, and confirmation state.
## Stage 2: Product
Instruct the agent to define information architecture, entities, workflows, roles, states, acceptance conditions, data lifecycle, deployment profile, dependency order, and cost implications. Preserve the ledger. Output product decisions and an updated spec.
## Stage 3: UI
Instruct the agent to collect only missing visual evidence, sanitize screenshot chrome and artifacts, and run the bundled `scripts/ui_design_intelligence/workbench_design.py` with the approved requirements and references. Generate one original, project-specific direction, persist the design system, and render desktop and phone prototypes. Enable light, dark, or system behavior only when required. Record explicit approval, or inherited approval when an existing approved UI has no material change. Do not begin implementation while the UI gate is unresolved.
## Stage 4: Development
Instruct the agent to implement a useful end-to-end vertical slice of the approved spec, create `workbench-manifest.json`, use the async storage and versioned backup contracts, use fictional removable demo data, keep secrets server-side, and record deviations. Output runnable code plus setup instructions and an updated control inventory.
## Stage 5: Testing
Instruct the agent to test every inventoried control plus persistence, migration, merge/overwrite import, corrupt input, mobile/desktop, overflow, accessibility, empty/error/offline, privacy, and recovery behavior appropriate to the level. Require expected behavior, actual results, screenshots or objective evidence, and known limitations. State that HTTP 200, syntax success, or selector existence alone does not prove functionality. Failed, blocked, or not-run required tests prevent `verification-passed`.
## Stage 6: Deployment
Instruct the agent to select deployment from device, sync, data-importance, privacy, and budget consequences. Explain the user/agent responsibility split, configure environment variables without exposing secrets, document data location/backup/restore/update/rollback/monitoring, and verify the selected runtime when authorized. A static URL or mobile-local package must not be described as cloud synchronization.
## Handoff checklist
Every stage ends with:
1. Files produced or changed.
2. Decisions added to `workbench-spec.md`.
3. Tests actually run and their results.
4. Open risks or questions.
5. Current workflow state and guard result.
6. Confirmation needed before the next stage.
7. A ready-to-paste context block for the next prompt.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |