Index of every page on this site: llms.txt
All skills
Run it yourself

Published skill from SDLC Skills Hub (SRS Architect themed). Extracted and structured for agent use.

  • 1 file
  • 10.1 KB
  • Updated last week
  • GitHub
Use this Skill: https://skilld.dev/gh/imawais-engineer/skills/sdlc-best-practices

Nothing lands on disk. Nothing to clean up.

Fork this Skill

Edit a local copy. It keeps the author and licence.

SKILL.md

≈30 tokens for metadata: the name and description. ≈2.5k when used: this file.

Description uses 1.5% of example budget

Before choosing a Skill, your Agent reads its name and description. All available Skills share that space.

  • A shorter description leaves more room for other Skills. This entry exceeds our 1% size suggestion.

In our Claude Code example, all Skill names and descriptions share 8,000 characters. This Skill uses ≈122 characters, or 1.5%.

The 1% threshold is a size suggestion. Longer descriptions can still fit.

Your model, settings, and other Skills decide how much text your Agent can read.

Example settings and source

The example uses a 200k-token context and default Claude Code settings. The count includes the name, description, separators, and when_to_use when present. Codex also counts local file paths.

Skit's source and limits: Codex 0.160.1, Claude Code 2.1.292.

V1.0.0 SDLC Best Practices

The gate skill. Forces a full discovery discussion and a complete SRS before any code is written, so the build starts from a specification instead of a guess.

SDLC Best Practices

Skill SK-01 · Phase 0 — Gate (Requirements) · Applies to: web apps, Android apps, APIs, AI apps The gate skill. No code is written until a complete, approved SRS exists.

  1. Purpose

The most expensive failure in AI-assisted development is building before the requirement is understood. A prompt such as "build me a booking app" is not a specification. When an AI builder codes from it, every gap is filled with a guess; each guess becomes a bug; each bug fix triggers another guess. That is the infinite edit loop.

This skill breaks the loop at its root. It forces a structured discovery, produces a Software Requirements Specification (SRS) in which every requirement has an ID and a testable acceptance criterion, and blocks implementation until a human approves it. Every other skill in this hub traces back to the requirement IDs created here.

  1. When to use — and when not to

Use it when:

Starting any new project (web, Android, API, AI app, internal tool). A change alters scope, data, roles, user flows, integrations, or platforms. An AI builder has needed more than two correction rounds on the same feature — that is a missing requirement, not a coding problem.

Do not use it for: a typo, a copy change, or a single-property style fix that is reversible in one edit.

  1. Inputs required Input Source Required Idea or brief in the user's own words User Yes Target users and who must be blocked User Yes Platform targets (web, Android, iOS, API) User / SK-02 Yes Existing systems, data, or code User / repository If any Constraints: deadline, budget, compliance, team skills User If any Interface direction (Interface Library prompt or "none") User Optional
  2. Procedure Step SDLC phase Action Output 1 Discovery Restate the problem in one sentence and ask the user to confirm it. Problem statement 2 Discovery Ask the discovery questions below, at most three per round, and wait. Never answer them yourself. Discovery notes 3 Analysis Run the completeness check (section 7). Every failed item becomes the next question. Gap list 4 Specification Write the SRS using the template in section 5. Number every requirement. SRS v0.x 5 Validation Run the ambiguity check: replace every vague word with a measurable rule. SRS v1.0 candidate 6 Approval Present the SRS and stop. Record the approval and date. Approved SRS v1.0 7 Planning Convert the SRS into a roadmap and tasks; every task names the FR IDs it satisfies. Roadmap, task list 8 Build Implement one task at a time, in roadmap order. Increment 9 Verify & validate After each task, run its acceptance criteria and record the result against the FR ID. Traceability matrix row 10 Change control Any new or changed behaviour updates the SRS first (new version), then the code. SRS changelog Discovery questions Area Ask Problem What breaks today? How is it handled now? What does success look like in one sentence? Users Who uses this? What can each role do that others cannot? Who must be blocked? Core flow Walk me through the single most important thing a user does, step by step. Data What is created, by whom, and what happens to it later? What must never be lost? Rules What is not allowed? Limits, approvals, pricing, eligibility, timing. Edge cases What happens on a duplicate, failure, cancellation, offline, or empty state? Platforms Web, Android, both? Offline? Installable? Minimum Android version or browsers? Out of scope What are we deliberately not building in version one? Ambiguity rule

These words are banned in requirements unless followed by a measurable definition: fast, simple, modern, clean, user-friendly, secure, scalable, intuitive, etc., some, many, manage, handle, support, appropriate, as needed.

✗ "The page loads fast." → ✓ "NFR-1: The dashboard renders first content in ≤ 2.0 s on a 4G connection (p75)." ✗ "Users can manage their profile." → ✓ "FR-7: A signed-in user can edit name (1–60 chars) and avatar (PNG/JPG ≤ 2 MB); email is read-only." 5. Deliverables Discovery notes — answers in the user's words. SRS — exactly this structure (IEEE 29148 aligned):

Software Requirements Specification — <Project> · v<version> · <date>

1. Purpose and problem

2. Scope

2.1 In scope (v1)

2.2 Out of scope (v1)

3. Stakeholders and user roles

Role Can do Cannot do Approx. count

4. Platforms and constraints

5. Functional requirements

ID Requirement (shall) Priority (Must/Should/Could) Acceptance criteria (Given / When / Then)
FR-1 Must Given … When … Then …

6. Non-functional requirements

ID Category Requirement Measure
NFR-1 Performance

7. Data model

Entity Fields (type, required) Owner Lifecycle states Access rule

8. User flows (numbered steps, including error and empty paths)

9. Edge cases

10. Assumptions (each confirmed or pending)

11. Open questions (blocking / non-blocking)

12. Roadmap

Phase Outcome FR IDs Depends on

13. Task list

  • T-1 — <task> — satisfies FR-1 — done when <observable outcome>

14. Traceability matrix

FR ID Task(s) Test ID(s) Status

15. Changelog

Roadmap and task list — inside the SRS (sections 12–13). Traceability matrix — inside the SRS (section 14), updated after every task. 6. Acceptance criteria Every functional requirement has a unique ID, a priority, and at least one Given/When/Then criterion. Every NFR has a measurable target. Every role states both what it can and cannot do. The out-of-scope list has at least one entry. No banned ambiguous word appears without a measurable definition. The human has explicitly approved the SRS version that the build uses. 7. Verification checklist (completeness check) Problem statement confirmed by the user. All roles listed, with "cannot do" filled in. Core flow written as numbered steps, including failure and empty paths. Every entity named with fields, owner, lifecycle, and access rule. Every FR has ID + priority + Given/When/Then. Every NFR has a number (time, size, count, percentage, version). Platforms stated (web / Android / both), with offline and minimum-version needs. Out of scope is explicit — silence is not scope. Every assumption is confirmed or listed as pending. Blocking open questions = 0. Roadmap and task list exist and every task cites an FR ID. Human approval recorded with SRS version and date. 8. Failure modes and anti-loop guards Failure mode Symptom Guard Infinite edit loop The same screen is regenerated 3+ times. Stop editing. Find the FR behind it; if none exists or it lacks Given/When/Then, write it, get approval, then make one edit that cites it. Building to impress A generic dashboard nobody asked for. Only build what an FR ID requires. Silent assumptions "Prices in USD" decided nowhere. Every guess goes to section 10 and must be confirmed. Scope by omission Everything is implicitly expected. Out-of-scope list is mandatory. Requirements as vibes "Manage profile" with no fields or rules. Ambiguity rule (section 4). Verifying by looking "It renders, so it's done." Done = acceptance criteria passed and recorded. Spec drift Behaviour changes mid-build; documents lie. Change control: SRS first, code second. "Just build it" User skips discovery. Reply with the smallest set of questions (≤ 3) that would make the build correct, and ask for one round of answers. Proceed only with an explicit waiver logged as assumptions. 9. Prompt template

Paste into any AI builder (Base44, Lovable, Cursor, Bolt, v0, Claude, ChatGPT) before any build request:

ROLE: You are a senior requirements engineer. You do not write code in this session. GOAL: Produce an approved SRS for: <one-sentence idea>. RULES:

  1. Ask at most 3 questions per round, grouped by area, each with concrete options. Wait for answers.

  2. Never invent answers. Anything you assume goes under "Assumptions" marked PENDING.

  3. Every functional requirement gets an ID (FR-n), a priority (Must/Should/Could), and Given/When/Then acceptance criteria.

  4. Every non-functional requirement (NFR-n) has a measurable target.

  5. Banned without a measurable definition: fast, simple, modern, clean, user-friendly, secure, scalable, manage, handle, support, etc.

  6. Include: scope, out of scope, roles (can / cannot), platforms, data model, user flows incl. error and empty paths, edge cases, open questions, roadmap, task list, traceability matrix.

  7. When the SRS is complete, print it and STOP. Do not generate code until I reply "APPROVED v1.0". CONTEXT I ALREADY HAVE: <paste notes, constraints, platform, interface direction>

  8. Traceability to the SRS

This skill creates the IDs every other skill uses:

FR-n → functional requirement · NFR-n → non-functional · T-n → task · TC-n → test case. Rule: every commit, build prompt, or AI edit request cites at least one FR or NFR ID. A change without an ID is a change request — update the SRS first. 11. Definition of done

The gate is passed when the completeness checklist is 100 % ticked, blocking open questions are zero, and the user has approved a specific SRS version. Building may start only then; the project is done only when every Must-priority FR in the traceability matrix shows Verified (implemented) and Validated (works as intended).

Related skills

Next: SK-02 Project Foundations & Architecture → SK-03 Complete App Blueprint → SK-06 Functional Prototype First. Use the SRS Architect in this hub to run this skill interactively.

RELATED LEARNING

Vibe Coding & AI-Assisted Development

28 MIN READ

Ready to hand this to your AI?

Copy the skill, drop it into your project, and require the AI to read it before the first prompt.

COPY SKILL DOWNLOAD .MD

Source: SKILL.md on GitHub

No rule matched.

skilld matched fixed text patterns in SKILL.md and file names. Patterns miss obfuscated code.

skilld run checks every file with the same patterns. It asks for approval before it loads a Skill with a behavior marked Needs approval.

No third-party reports yet.

Signed by skilld at 5ecc5ec. This ties the file your Agent reads to that commit on GitHub. It does not review the instructions.

Last checked against GitHub 5 days ago.

Activeupdated last week
source
https://sdlc.base44.app/skill/sdlc-best-practices
priority
foundation-first

README badge

README badge for imawais-engineer/skills/sdlc-best-practices