Kickoff Prompt

Prompt sections

Every section is optional.

Included

Skipped

INCLUDE

Kickoff Prompt

29 included sections - 67/100 quality score

Project name

Template

Step 1 of 29 - Agent role3%

Current section

Agent role

Define the neutral role, working style and decision rules the coding agent should follow. This keeps the prompt model-neutral while setting expectations for how work should be approached.

INCLUDE

Role

Empty values render as not specified.

Working style

Empty values render as not specified.

AI assist

Uses the AI connection saved in AI settings. Credentials never enter the generated prompt.

Prompt quality

Deterministic score, no AI call required.

67/100

Good draft31 unspecified - 0 skipped - 100% content depth

Strengths

Objective section contains usable direction.
Acceptance criteria are present for verification.
The output contract says what the agent must return.
Operating instructions define how the agent should work.

Weaknesses

Acceptance criteria are not measurable - add checkable conditions.
31 included fields still render as "not specified".
Concrete source or reference material is not specified yet.

Missing fields

Project context: Context
Target users: Use cases
Business goal: Business goal
Business goal: KPIs
Project status: Status
Project status: Existing assets
Source references: Repository references
Source references: External references
Source references: Sample inputs and data
Frontend: Frontend requirements
Backend: Backend requirements
API: API contracts

Suggestions

Fill the highest-impact missing fields: objective, features, acceptance criteria and verification commands.
Add repo paths, design links, docs or sample data so the agent can ground its work.
Specify runtime, Docker and deployment expectations so implementation choices are testable.
Add the minimum signals worth observing after launch.

Final output

Ready to copy or download. Exports include only INCLUDE sections.

Agent files:

Omit unspecified optional fields

# Kickoff Prompt

> **Kickoff brief for an AI coding agent. Read the whole brief before writing any code.**
> - Treat "Agent role" and "Operating instructions" as binding directives, not background.
> - Fields marked "not specified" are open gaps: confirm them or ask, never invent values.
> - Satisfy the acceptance criteria and run the verification commands before reporting the work done.

## Agent role

### Role
Act as a senior full-stack software engineer who ships production-ready, well-tested code.

### Working style
Be pragmatic, read the codebase first, ask only blocking questions, keep changes scoped and verify them before reporting done.

## Project objective

### Objective
Build a chat assistant that answers questions grounded on a private knowledge base using retrieval-augmented generation.

### Success outcome
Users get accurate, cited answers from their documents; out-of-scope questions are handled gracefully.

## Project context

### Domain
AI / knowledge assistant

## Target users

### Users
Support agents, employees or customers querying internal knowledge.

## Source references

### Repository references
not specified

### External references
not specified

### Sample inputs and data
not specified

## Features

### Core features
Document ingestion, chunking & embeddings, semantic search, chat UI with streaming and citations, conversation history.

### Out of scope
Fine-tuning models and autonomous agent actions.

## Frontend

### Frontend stack
Next.js, React, TypeScript, Tailwind CSS, streaming chat UI

## Backend

### Backend stack
Next.js Route Handlers, an embeddings pipeline and a vector store client

## Database

### Database stack
PostgreSQL with pgvector (or a managed vector DB)

### Data model
Document, Chunk(embedding), Conversation, Message.

## API

### API style
REST + server-sent events for token streaming

## Security

### Security requirements
Scope retrieval to the user's documents; never leak other tenants' content; do not log prompts containing secrets.

### Sensitive data
Uploaded documents and conversation content.

## Team perspectives

### Product perspective
Review whether the main workflow solves the stated user problem and flag scope creep before implementing optional features.

### Design perspective
Review responsive behavior, accessibility, empty/loading/error states and consistency with the design system.

### Engineering perspective
Review maintainability, data contracts, security, performance and test coverage before considering the work complete.

## Time, budget and tradeoffs

### Priority tradeoffs
Prioritize the core workflow and verifiable correctness before polish, automation or optional integrations.

## Acceptance criteria

### Criteria
Answers cite sources, respect document scope, and stream smoothly.

## Verification commands

### Commands
not specified

## Output contract

### Final response format
Return a concise final response with: completed work, changed files, verification results, known gaps and recommended next steps.

### Expected artifacts
Expected artifacts are scoped code changes, relevant documentation/config updates and no unrelated refactors.

### Verification evidence
Report every verification command run. If a check cannot run, explain the blocker and residual risk.

## Operating instructions

### Instructions
Read the existing context first, keep changes scoped, do not invent missing requirements and verify before reporting completion.

### Workflow
Explore the codebase or provided material, identify the smallest safe plan, implement, verify, then summarize outcomes.

### Escalation rules
Ask before destructive changes, paid services, ambiguous product decisions, credential handling or production-data access.