AI-SDLC Agent Skill
Open-source agent skill standardizing Socratic intent, 9 specialist roles, and multi-harness development.
Creator & Lead Maintainer
Open Source
2026
Active Development
Executive Summary & Strategic Overview
Open-source agent skill on skills.sh that turns any repository into an autonomous development environment with 9 specialist agent roles, Socratic intent capture, and cross-editor bridge adapters.
Verified Performance Metrics & Business Impact
Open Registry
One-command installable agent skill via npx skills add jedmamosto/ai-sdlc
Specialist Fleet
Complete SDLC role coverage from product strategy to code reviewer
Test Driven
16 automated test suites with PreToolUse test tampering protection
Conflict-Free Work
Subsystem write mutex locks and 5-stage pipeline for coordinated teamwork
The Engineering Challenge
Deploying AI coding agents without structure causes scope creep, unhandled UI edge states, concurrent file overwrite collisions, and fragmented instructions across different editors.
The Architectural Solution
Built an installable open-source agent skill providing Socratic intent intake, 9 specialized agent personas, and a 5-stage Hub-and-Spoke workflow with write mutex locks and multi-harness adapters.
Core Technical & Architectural Highlights
Socratic Intake & Problem Space Triad
Enforces rule-of-one conversational intake to isolate the core user problem before the Stage 0 strategy triad defines UX flows, wireframes, and architecture.
9 Specialized Agent Personas
Establishes clear division of labor across engineering management, UX research, UI art direction, system architecture, frontend, backend, QA, and documentation.
Controlled Language & Pre-Tool Protection
Applies ASD-STE100 plain English prompt rules to eliminate hallucinations, paired with PreToolUse hooks guarding test files against tampering during refactors.
Agentic AI Pair Programming & Antigravity Acceleration
Engineered with test-driven development in Antigravity, verifying workspace initializers, agent fleet seeders, and hook handlers across 16 automated test suites with zero failures.