ReddRadar
Agent skill
preview
Display execution plan preview before running commands, allowing user confirmation or modification.
Stars
163
Forks
31
Install this agent skill to your Project
npx add-skill https://github.com/majiayu000/claude-skill-registry/tree/main/skills/data/preview-benjaminrose805-react-basecamp
SKILL.md
Preview Skill
Display execution plan preview before running commands, allowing user confirmation or modification.
Mode Check
CRITICAL: Check mode before displaying preview.
typescript
if (mode === "basic") {
// Skip preview entirely
// Execute command immediately
return;
}
// Continue with preview display
When Used
| Command | Preview Content |
|---|---|
/start |
Worktree path, branch name, next steps |
/plan |
Mode (define/reconcile), phases, sub-agents |
/implement |
Spec reference, stages, agents, TDD phases |
/ship |
Commit, PR, CI, CodeRabbit stages |
NOT used by: /guide (informational), /mode (immediate switch)
Preview Data Structure
typescript
interface ExecutionPreview {
// Command context
command: string; // Original command (e.g., "/implement login form")
detected: string; // What was detected (e.g., "Frontend component")
scope: string; // Scope description (e.g., "Login form UI")
// Routing info
route: {
type: "agent" | "workflow";
name: string; // Agent or workflow name
reason: string; // Why this route was chosen
};
// Execution plan
stages: Stage[];
// Metadata
tools: string[]; // MCP servers/CLI tools that will be used
flags?: string[]; // Any flags that modify behavior
}
interface Stage {
number: number;
name: string; // e.g., "RESEARCH", "BUILD", "VERIFY"
agent?: string; // Agent handling this stage
subAgents: SubAgentPlan[];
status: "pending" | "conditional"; // conditional = depends on prior stage
}
interface SubAgentPlan {
name: string; // e.g., "ui-researcher"
model: "Opus" | "Sonnet" | "Haiku";
tasks: string[]; // Planned tasks
parallel?: boolean; // True if runs in parallel with siblings
}
Display Formats
/start Preview
text
┌─────────────────────────────────────────────────────────────────┐
│ /start user-authentication │
├─────────────────────────────────────────────────────────────────┤
│ Creating new workspace for: user-authentication │
│ │
│ ACTIONS │
│ ┌─────────────────────────────────────────────────────────────┐│
│ │ 1. CREATE WORKTREE ││
│ │ Path: ../project-user-authentication ││
│ │ Branch: feature/user-authentication ││
│ │ ││
│ │ 2. NEXT STEPS ││
│ │ → Restart session in new worktree ││
│ │ → Run /plan to begin designing ││
│ └─────────────────────────────────────────────────────────────┘│
│ │
│ [Enter] Run [e] Edit name [Esc] Cancel │
└─────────────────────────────────────────────────────────────────┘
/plan Preview (Define Mode)
text
┌─────────────────────────────────────────────────────────────────┐
│ /plan user-authentication │
├─────────────────────────────────────────────────────────────────┤
│ Mode: Define │
│ Feature: User authentication with email/password │
│ │
│ PHASES │
│ ┌─────────────────────────────────────────────────────────────┐│
│ │ 1. RESEARCH domain-researcher Opus ││
│ │ □ Search existing auth patterns ││
│ │ □ Check for conflicts ││
│ │ □ Identify integration points ││
│ ├─────────────────────────────────────────────────────────────┤│
│ │ 2. WRITE domain-writer Sonnet ││
│ │ □ Create requirements.md (EARS format) ││
│ │ □ Create design.md (architecture) ││
│ │ □ Create tasks.md (phased work items) ││
│ ├─────────────────────────────────────────────────────────────┤│
│ │ 3. VALIDATE quality-validator Haiku ││
│ │ □ Verify EARS compliance ││
│ │ □ Check acceptance criteria ││
│ └─────────────────────────────────────────────────────────────┘│
│ │
│ Output: specs/user-authentication/ │
│ │
│ [Enter] Run [e] Edit [?] Details [Esc] Cancel │
└─────────────────────────────────────────────────────────────────┘
/plan Preview (Reconcile Mode)
text
┌─────────────────────────────────────────────────────────────────┐
│ /plan (reconcile PR #42) │
├─────────────────────────────────────────────────────────────────┤
│ Mode: Reconcile │
│ PR: #42 - Add user authentication │
│ CodeRabbit comments: 3 │
│ │
│ ISSUES TO ADDRESS │
│ ┌─────────────────────────────────────────────────────────────┐│
│ │ 1. [Security] Use bcrypt, not SHA256 ││
│ │ 2. [Performance] Add index on email column ││
│ │ 3. [Style] Use early returns in validatePassword() ││
│ └─────────────────────────────────────────────────────────────┘│
│ │
│ PHASES │
│ ┌─────────────────────────────────────────────────────────────┐│
│ │ 1. ANALYZE domain-researcher Opus ││
│ │ □ Review each CodeRabbit comment ││
│ │ □ Identify affected files ││
│ │ □ Assess fix complexity ││
│ ├─────────────────────────────────────────────────────────────┤│
│ │ 2. PLAN domain-writer Sonnet ││
│ │ □ Create fix plan with tasks ││
│ │ □ Prioritize by severity ││
│ └─────────────────────────────────────────────────────────────┘│
│ │
│ [Enter] Run [e] Edit [?] Details [Esc] Cancel │
└─────────────────────────────────────────────────────────────────┘
/implement Preview
text
┌─────────────────────────────────────────────────────────────────┐
│ /implement user-authentication │
├─────────────────────────────────────────────────────────────────┤
│ Spec: specs/user-authentication/ (approved) │
│ Tasks: 12 across 4 phases │
│ TDD: Enabled (red → green → refactor) │
│ │
│ STAGE 1: DATABASE SCHEMA │
│ ┌─────────────────────────────────────────────────────────────┐│
│ │ Agent: code-agent ││
│ │ ││
│ │ 1. RESEARCH code-researcher Opus ││
│ │ □ Find existing DB patterns ││
│ │ 2. TDD-RED code-writer Sonnet ││
│ │ □ Write failing tests ││
│ │ 3. TDD-GREEN code-writer Sonnet ││
│ │ □ Implement to pass tests ││
│ │ 4. VALIDATE code-validator Haiku ││
│ │ □ Verify tests pass ││
│ └─────────────────────────────────────────────────────────────┘│
│ │
│ STAGE 2: AUTH API │
│ └─ [Same pattern with code-agent] │
│ │
│ STAGE 3: UI COMPONENTS │
│ └─ [Same pattern with ui-agent] │
│ │
│ STAGE 4: FINAL VERIFICATION │
│ ┌─────────────────────────────────────────────────────────────┐│
│ │ Agent: check-agent (parallel) Haiku ││
│ │ ⊕ build-checker ││
│ │ ⊕ type-checker ││
│ │ ⊕ lint-checker ││
│ │ ⊕ test-runner ││
│ │ ⊕ security-scanner ││
│ └─────────────────────────────────────────────────────────────┘│
│ │
│ Tools: cclsp, context7, next-devtools │
│ │
│ [Enter] Run [e] Edit [?] Details [Esc] Cancel │
└─────────────────────────────────────────────────────────────────┘
/ship Preview
text
┌─────────────────────────────────────────────────────────────────┐
│ /ship │
├─────────────────────────────────────────────────────────────────┤
│ Feature: user-authentication │
│ Branch: feature/user-authentication │
│ Changes: 8 files, +342 -12 │
│ │
│ STAGES │
│ ┌─────────────────────────────────────────────────────────────┐│
│ │ 1. COMMIT ││
│ │ Agent: git-agent ││
│ │ ├─ change-analyzer (Sonnet) - Generate message ││
│ │ └─ git-executor (Haiku) - Create commit ││
│ ├─────────────────────────────────────────────────────────────┤│
│ │ 2. CREATE PR ││
│ │ Agent: git-agent ││
│ │ ├─ pr-analyzer (Sonnet) - Generate description ││
│ │ └─ git-executor (Haiku) - Create PR via gh CLI ││
│ ├─────────────────────────────────────────────────────────────┤│
│ │ 3. WAIT FOR CI ││
│ │ □ Monitor GitHub Actions ││
│ │ □ Report pass/fail ││
│ ├─────────────────────────────────────────────────────────────┤│
│ │ 4. WAIT FOR CODERABBIT ││
│ │ □ Monitor CodeRabbit review ││
│ │ □ Report comments or approval ││
│ └─────────────────────────────────────────────────────────────┘│
│ │
│ [Enter] Run [e] Edit [Esc] Cancel │
└─────────────────────────────────────────────────────────────────┘
User Interactions
Keyboard Controls
| Key | Action | Description |
|---|---|---|
Enter |
Run | Execute the plan as shown |
e |
Edit | Modify scope or options |
? |
Details | Show expanded details (tools, context) |
Esc |
Cancel | Abort without executing |
Edit Mode
When user presses e:
text
┌─────────────────────────────────────────────────────────────────┐
│ Edit Execution Plan │
├─────────────────────────────────────────────────────────────────┤
│ │
│ Scope: login form │
│ ───────────────── │
│ Current: "Login form UI" │
│ New scope: _ │
│ │
│ Options: │
│ ───────── │
│ [ ] Skip research phase │
│ [ ] Skip validation phase │
│ [x] Include accessibility audit │
│ │
│ [Enter] Apply [Esc] Cancel │
│ │
└─────────────────────────────────────────────────────────────────┘
Details View
When user presses ?:
text
┌─────────────────────────────────────────────────────────────────┐
│ Execution Details │
├─────────────────────────────────────────────────────────────────┤
│ │
│ MCP Servers: │
│ ───────────── │
│ • cclsp - TypeScript LSP for code navigation │
│ • shadcn - Component registry lookup │
│ • playwright - E2E and accessibility testing │
│ • context7 - Library documentation lookup │
│ │
│ CLI Tools: │
│ ────────── │
│ • pnpm test - Run Vitest tests │
│ • pnpm typecheck - TypeScript checking │
│ • pnpm lint - ESLint checking │
│ │
│ Spec Reference: │
│ ─────────────── │
│ • specs/login-form/design.md │
│ │
│ Estimated Context Usage: │
│ ───────────────────────── │
│ • Research: ~15% (Opus) │
│ • Build: ~50% (Sonnet) │
│ • Validate: ~10% (Haiku) │
│ │
│ [Any key] Back │
│ │
└─────────────────────────────────────────────────────────────────┘
Automation Flag
Skip Preview with --yes
bash
/implement --yes # Execute immediately without preview
/ship --yes # Ship without confirmation
/start auth --yes # Create worktree without preview
When --yes flag is present:
- Skip preview display
- Use default options
- Execute immediately
- Still show progress during execution
When to Use --yes
| Scenario | Recommended |
|---|---|
| CI/CD automation | Yes |
| Scripted workflows | Yes |
| Repetitive tasks | Yes |
| New/unfamiliar tasks | No |
| Complex features | No |
| Production changes | No |
Preview Generation
Algorithm
typescript
async function generatePreview(
command: string,
args: string
): Promise<ExecutionPreview> {
// 1. Parse command
const { type, target, flags } = parseCommand(command, args);
// 2. Route command
const routing = await route(type, target);
// 3. Build preview based on route type
if (routing.type === "agent") {
return buildAgentPreview(routing.agent, target);
} else {
return buildWorkflowPreview(routing.workflow, target);
}
}
function buildAgentPreview(agent: Agent, target: string): ExecutionPreview {
return {
command: `/${agent.command} ${target}`,
detected: agent.detectType(target),
scope: target,
route: {
type: "agent",
name: agent.name,
reason: agent.routingReason,
},
stages: agent.phases.map((phase) => ({
number: phase.order,
name: phase.name,
agent: agent.name,
subAgents: [
{
name: phase.subAgent,
model: phase.model,
tasks: phase.defaultTasks,
},
],
status: "pending",
})),
tools: agent.tools,
};
}
Error Handling
| Scenario | Handling |
|---|---|
| Routing fails | Show error, suggest alternatives |
| Spec not found | Show warning, proceed with keyword routing |
| Invalid edit input | Show validation error, re-prompt |
| User cancels | Exit gracefully, no execution |
| Preview generation fail | Fall back to simple text description |
Output
Preview Accepted
markdown
## Preview: ACCEPTED
Executing plan as shown...
(Switches to progress display)
Preview Edited
markdown
## Preview: MODIFIED
Changes:
- Scope: "login form" → "login form with 2FA"
- Skipped: research phase
Executing modified plan...
Preview Cancelled
markdown
## Preview: CANCELLED
No changes made.
Didn't find tool you were looking for?