تشغيل أي مهارة في Manus بنقرة واحدة

adlc

النجوم٤

التفرعات٠

آخر تحديث٣ يونيو ٢٠٢٦ في ٢١:٢٤

Full Agent Development Life Cycle pipeline for the aiagent project. Runs story refinement → design → implementation planning (with human approval checkpoint) → implementation → testing → security → database → devops. Each coding phase uses a doer + antagonistic reviewer loop requiring score >= 8/10.

التثبيت

التثبيت باستخدام Codex أو Claude انسخ هذا Prompt والصقه في Codex أو Claude أو مساعد آخر ليراجع صفحة Skill ويثبّتها لك.

تشغيل في Manus

المصدر

drujensen

drujensen/aiagent

فتح مستودع GitHub عرض مستودعات المنشئ

تنزيل

تشغيل في Manus

المهن ذات الصلةSOC

استنادا إلى تصنيف SOC المهني

مطوّرو البرمجياتمهن الحاسوب والرياضيات·SOC 15-1252

SKILL.md

readonly

name	adlc
description	Full Agent Development Life Cycle pipeline for the aiagent project. Runs story refinement → design → implementation planning (with human approval checkpoint) → implementation → testing → security → database → devops. Each coding phase uses a doer + antagonistic reviewer loop requiring score >= 8/10.
user-invocable	true
argument-hint	add a new tool that searches GitHub repositories
context	fork

ADLC Pipeline

Run the full development pipeline for:

Stage 1: Pre-Coding (no code is written until the plan is approved)

Step 1: Story Refinement (interactive)

Use the Agent tool with subagent_type: "story-refiner" to:

Explore the relevant aiagent codebase
Generate a focused batch of 3–6 clarifying questions
Present the questions to the user and collect answers
Repeat until the story is complete (all unknowns resolved)

Capture and store the full story-refiner output (both Refined Story and Relevant Codebase Context sections). This is passed verbatim to every subsequent agent.

When complete, ask: "Story looks complete — ready to design?"

Step 2: Design (review loop)

Use the Agent tool with subagent_type: "architect" to design the solution. Pass it:

The full story-refiner output (Refined Story + Relevant Codebase Context)

The architect defaults to maintaining existing DDD patterns. Deviations are explicitly justified.

Then use the Agent tool with subagent_type: "architect-reviewer" to review the design. Pass the architect's full output.

Iterate up to 3 times until score >= 8/10 (APPROVED) or escalate to user.

Capture and store the approved design document — passed verbatim to the planner.

Step 3: Implementation Plan (review loop + human checkpoint)

Use the Agent tool with subagent_type: "planner" to produce an implementation plan. Pass it:

The full story-refiner output
The approved design document

The planner produces a Reuse Inventory and numbered implementation steps.

Then use the Agent tool with subagent_type: "planner-reviewer" to verify the plan.

Iterate up to 3 times until score >= 8/10.

PAUSE. Present the complete plan to the user (Reuse Inventory + all numbered steps).

Ask explicitly: "Does this plan look correct? Type 'proceed' to begin coding, or describe any changes needed."

Do NOT advance to Stage 2 until the user types 'proceed' or equivalent confirmation.

If the user requests changes, update the plan and re-present it.

Capture and store the approved plan — passed verbatim to the developer.

Stage 2: Coding (only after human approves the plan)

For each coding phase, pass to the agent: the approved plan (including Reuse Inventory), the approved design, and the refined story (including acceptance criteria).

Step 4: Implementation

Doer: Use the Agent tool with subagent_type: "developer"

Pass:

Approved plan (Reuse Inventory + numbered steps)
Approved design document
Refined story (acceptance criteria)

The developer will:

Establish a green test baseline first
Execute one plan step at a time with test gates
Run full QA when complete: go fmt ./... && go vet ./... && go mod tidy && go build . && go test ./... -race

Reviewer: Use the Agent tool with subagent_type: "developer-reviewer"