Srs To Delivery Plan

SRS to Delivery Plan

Convert a Software Requirements Specification (SRS) into a structured delivery plan:

SRS → EPICS → STORIES → TASKS → SPRINT PLAN

This system combines:

• Product thinking (EPICS)
• Vertical slicing (STORIES)
• Execution clarity (TASKS)
• Delivery planning (SPRINTS)

---

Core Principles

Vertical Slice (Tracer Bullet)

Each STORY must:

• Deliver a thin but complete end-to-end flow
• Cut through ALL layers (DB → API → UI → integration → tests)
• Be independently testable and demoable
• Provide real user value

Avoid horizontal slices like:

• "Build database schema"
• "Create API layer"

Instead:

• "User can create account and log in"

---

HITL vs AFK

HITL (Human-in-the-loop)

Requires:

• Product decisions
• Architecture discussions
• UX validation

AFK

• Can be implemented and merged independently
• No external decision required

Always maximize AFK stories.

---

PROCESS

---

Step 1 — Locate the SRS

Ask the user for:

• GitHub issue number OR
• URL OR
• Raw SRS document

If not available → ask for it.

Do NOT proceed without understanding the SRS.

---

Step 2 — Understand the System

Extract:

• Core features
• User stories
• Key flows
• System boundaries
• Constraints (tech, product, compliance)

Summarize briefly before proceeding.

---

Step 3 — Create EPICS

Group features into high-level capabilities.

EPIC Rules

• Represents a major deliverable
• Takes ~1–3 weeks
• Contains multiple STORIES
• Focus on outcomes, not implementation

Output Format

For each EPIC:

• Title
• Description
• Covered SRS sections

---

Step 4 — Create STORIES (Vertical Slices)

Break each EPIC into STORIES.

STORY Rules

• End-to-end functionality
• Demoable independently
• Small (ideally 1–2 days)
• User-value focused

STORY Fields

• Title
• Type: HITL / AFK
• Blocked by
• User stories covered (from SRS)

---

Step 5 — Create TASKS

Break STORIES into TASKS. Each task must be a self-contained work packet with enough detail that the implementer (human or AI agent) can execute without further clarification.

TASK Rules

• Technical units of work
• 2–8 hours each
• Can be layer-specific
• NOT user-facing
• Must be independently executable with full context

Required TASK Fields

Every task MUST include ALL of the following:

| Field | Description | |-------|-------------| | Context | Why this task exists. What story/feature it enables. What came before it and what comes after | | Problem Statement | The specific technical gap this task fills. What doesn't exist or work yet | | Scope | Exactly what files/modules are touched. What is IN scope and what is explicitly OUT of scope | | Definition of Done | Concrete, binary checklist — when is this task finished? No ambiguity | | Acceptance Criteria | Behavioral criteria from the user/product perspective that this task must satisfy | | Testing Plan | What tests to write or run. Unit tests, integration tests, manual verification steps | | Validation | How to verify the task is complete — specific commands to run, pages to check, DB queries to inspect | | Key Files | Existing files to read/modify, with brief explanation of what each provides | | Implementation Notes | Patterns to follow, functions to reuse, gotchas to avoid. Reference existing code by path | | Framework Integration | (Required if task adds a third-party library) SSR compatibility, Next.js config flags, React version requirements, required wrappers or polyfills. Read the library's framework docs and document findings here |

TASK Examples

• Define DB schema
• Implement API endpoint
• Build UI component
• Write integration test

---

Step 6 — Quiz the User (MANDATORY)

Present:

EPICS

STORIES (grouped by EPIC)

TASKS (optional preview)

Ask:

• Is EPIC grouping correct?
• Are stories too big or too small?
• Any missing flows?
• Are dependencies valid?
• Are HITL vs AFK classifications correct?
• Should any stories be split or merged?

Iterate until user approves.

---

Step 7 — Create GitHub Issues

Create issues in this order:

1. EPICS
2. STORIES
3. (Optional) TASK issues

Use the issue templates from [template.md](template.md) for EPIC, STORY, and TASK formatting.

TASK Handling

Two options:

Option A (default)

Keep tasks as a checklist inside the STORY issue.

Option B (advanced teams)

Create sub-issues per task. Use the TASK template from [template.md](template.md).

---

Step 8 — Sprint Planning

After STORIES are created, organize them into sprints.

Sprint Rules

• Duration: 1 week (recommended)
• Prioritize:

1. Unblocking stories
2. Core flows
3. First usable system

• Balance workload based on capacity

Sprint Output Format

# Sprint Plan

## Sprint 1
Goal: First working end-to-end flow

Stories:
- #12 User signup & login (AFK)
- #13 Basic dashboard (AFK)
- #14 First data capture (HITL)

---

## Sprint 2
Goal: Core functionality expansion

Stories:
- #15 Data processing (AFK)
- #16 Task extraction (AFK)

Prioritization Strategy

1. Foundations (auth, schema, infra)
2. First usable loop
3. Feature expansion
4. Optimization & scale

---

Rules to Enforce

• Do NOT skip vertical slicing
• Do NOT create horizontal-only stories
• Do NOT over-bundle stories
• Prefer many small stories over few large ones
• Always validate with user before creating issues
• Always maintain dependency correctness

---

Final Output Checklist

Before finishing, ensure:

• [ ] EPICS clearly defined
• [ ] STORIES follow vertical slicing
• [ ] TASKS are actionable
• [ ] Dependencies are valid
• [ ] HITL vs AFK correctly assigned
• [ ] Sprint plan is realistic
• [ ] User has approved structure

---

Optional Enhancements

If user asks, also provide:

• Jira version
• Linear version
• Confluence documentation
• Roadmap view (monthly)
• Team allocation plan

---

Example (Reference)

EPIC: Memory System

Stories:

1. Save journal entry (AFK)
2. Extract structured memory (AFK)
3. Link memory to entities (HITL)

---

This skill transforms a static SRS into a complete execution-ready system.

See [template.md](template.md) for all issue templates.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: Rijul1204
• Source: Rijul1204/rashedul-agentic-engineering
• License: MIT

Install and usage instructions live in the source repository linked above.