Back to list
5dlabs

prd-analysis

by 5dlabs

Cognitive Task Orchestrator - GitOps on Bare Metal or Cloud for AI Agents

2🍴 1📅 Jan 24, 2026

SKILL.md


name: prd-analysis description: PRD parsing and task decomposition patterns for intake workflows. agents: [morgan] triggers: [prd, intake, task decomposition, planning, project management]

PRD Analysis and Task Decomposition

Patterns for parsing PRDs and generating comprehensive task documentation.

Critical Success Criteria

Output quality is measured by how well agents can implement tasks WITHOUT needing to re-read the PRD. Each task must be self-contained with all necessary context embedded.

Structured Analysis (Chain of Thought)

Before generating tasks, think step-by-step:

  1. Analyze PRD Structure - Identify features, tech requirements, implicit dependencies
  2. Map Dependencies - What depends on what? Shared infrastructure needs?
  3. Define Task Boundaries - Atomic, single responsibility, independently testable
  4. Generate Implementation Details - Pseudo-code, file structures, library versions
  5. Define Test Strategies - Acceptance criteria, unit/integration/E2E guidance
  6. Self-Verify - All requirements covered? No circular deps? Valid ordering?

Service Discovery

Before generating any tasks, analyze the PRD to identify all services and their tech stacks.

Create a Service-to-Agent Mapping Table:

Service NameAgentTech StackDescription
Notification RouterrexRust/AxumHigh-performance API
Integration ServicenovaBun/Elysia/EffectChannel delivery
Admin APIgrizzGo/gRPCTenant management
Web ConsoleblazeNext.js/React/shadcnAdmin dashboard
Mobile ApptapExpo/React NativeMobile client
Desktop ClientsparkElectronDesktop notifications

Agent Assignment Rules

Implementation Agents (Write Code)

AgentLanguage/StackUse For
boltKubernetes/HelmInfrastructure (Task 1 ONLY)
rexRust/Axum/TokioRust backend services
grizzGo/gRPC/ChiGo backend services
novaBun/Elysia/EffectTypeScript backend
blazeNext.js/React/shadcnWeb frontends
tapExpo/React NativeMobile apps
sparkElectronDesktop apps

Support Agents (Review Only - AFTER Implementation)

AgentRoleUse ONLY For
cleoQuality ReviewCode review AFTER PR
cipherSecurity AuditSecurity audit AFTER PR
tessTestingWriting tests AFTER PR
atlasIntegrationMerging PRs AFTER reviews

CRITICAL: Support agents are NEVER assigned to implementation tasks.

Task 1: Always Infrastructure

Task 1 must ALWAYS be assigned to Bolt to provision:

  • Databases (PostgreSQL, MongoDB)
  • Caches (Redis/Valkey)
  • Message queues (Kafka, NATS)
  • Object storage (SeaweedFS S3)

Code Signatures in Details

Include language-specific function/struct signatures in task details:

Rust Example

#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct Notification {
    pub id: Uuid,
    pub tenant_id: Uuid,
    pub channel: Channel,
}

pub async fn create_notification(
    State(state): State<AppState>,
    Json(req): Json<CreateNotificationRequest>,
) -> Result<Json<NotificationResponse>, ApiError>

TypeScript/Effect Example

import { Schema } from "@effect/schema"
import { Effect } from "effect"

export const Integration = Schema.Struct({
  id: Schema.UUID,
  name: Schema.String,
  channel: Schema.Literal("slack", "discord", "email"),
})

Go Example

type TenantServiceServer struct {
    pb.UnimplementedTenantServiceServer
    db *pgxpool.Pool
}

func (s *TenantServiceServer) CreateTenant(
    ctx context.Context,
    req *pb.CreateTenantRequest,
) (*pb.Tenant, error)

Task JSON Format

{
  "tasks": [
    {
      "id": "1",
      "title": "Provision Infrastructure",
      "description": "Deploy PostgreSQL, Redis, Kafka clusters",
      "priority": "critical",
      "status": "pending",
      "dependencies": [],
      "agentHint": "bolt",
      "details": "Full implementation details with code signatures",
      "testStrategy": "Verification commands and expected outputs"
    }
  ]
}

Quality Checklist

Before finalizing:

  • Service table created with all services from PRD
  • Task 1 is agentHint: "bolt" for infrastructure
  • Every task has agentHint set
  • Implementation agents only for code tasks
  • Support agents ONLY for review/audit tasks
  • Code signatures in details for all implementation tasks
  • No auth/jwt/oauth tasks assigned to cipher
  • Dependencies form valid DAG (no cycles)
  • Implementation tasks depend on task-1
  • Each task is atomic (completable in one PR)

Score

Total Score

65/100

Based on repository quality metrics

SKILL.md

SKILL.mdファイルが含まれている

+20
LICENSE

ライセンスが設定されている

+10
説明文

100文字以上の説明がある

0/10
人気

GitHub Stars 100以上

0/15
最近の活動

1ヶ月以内に更新

+10
フォーク

10回以上フォークされている

0/5
Issue管理

オープンIssueが50未満

+5
言語

プログラミング言語が設定されている

+5
タグ

1つ以上のタグが設定されている

+5

Reviews

💬

Reviews coming soon