⚡ v2.1.0 — AI Pair Programming Mode

Stop Prompting.
Start Specifying.

Structural gates that block AI agents from breaking architecture. Not warnings. Not suggestions. Enforcement.

11
Structural Gates
10
Workflow Commands
11
Agent Personas
30+
Frameworks Detected

AI Agents Are Fast. But Undisciplined.

Speed without structure is organized chaos. Here's what happens when AI agents run free.

🔮

Hallucinated APIs

The AI calls functions that don't exist in your codebase, creating phantom dependencies.

🔀

Layer Skipping

React components importing directly from database adapters. Architecture? Ignored.

🔑

Secret Leaks

API keys hardcoded in utility files. One commit away from exposure.

📐

Spec Drift

You asked for a login page. You got social auth, password strength meter, and SSO — unrequested.

🏗️

Over-Engineering

A simple CRUD endpoint becomes a 4-layer abstraction with a factory pattern.

💬

Prompt Degradation

By message 50, your system prompt is a distant memory. The AI drifts further each turn.

Prompts Request. Structure Enforces.

CLAUDE.md and Cursor Rules are suggestions. Harness gates are enforcement.

❌ Prompt-based (CLAUDE.md / Rules)

  • AI acknowledges rules, then violates them
  • Degrades over long conversations
  • No memory of past violations
  • Same issues repeat every session
  • Suggestions, not enforcement

✓ Structure-based (Harness)

  • Gates block file writes that violate rules
  • Fires on every write, regardless of context length
  • Every violation becomes a permanent rule
  • System gets stronger with use
  • Enforcement, not suggestion

Gates That Block, Not Warn

PostToolUse hooks fire on every file write. The AI literally cannot save a file that violates the rules.

Gate What It Catches Status
check-secrets35+ secret patterns (AWS, GitHub, Stripe, OpenAI, …)Default
check-boundariesUnauthorized imports and dependenciesDefault
check-structureFiles in wrong directoriesDefault
check-specIncomplete specs (no TODO/TBD allowed)Default
check-layers3-tier violations (P→D skip, reverse deps)Default
check-securitySAST scan (Semgrep / Bandit / built-in)Default
check-depsKnown vulnerability auditDefault
check-complexityCyclomatic complexity thresholdOpt-in
check-mutationMutation testing coverageOpt-in
check-performancePerformance regressionOpt-in
check-ai-antipatternsCommon AI code generation mistakesOpt-in

Specify First. Code Second.

A Socratic interview measures ambiguity across 4 dimensions before a single line of code is written.

/interview → /seed → /trd → /decompose → /run → /evaluate
     ↑                                                  │
     └──────── /evolve (until convergence) ────────┘
🎤

/interview

Socratic interview measuring 4D ambiguity — goal, constraints, success criteria, context.

🌱

/seed

Crystallize into an immutable seed spec. Changes mean new versions, never mutations.

🔨

/run

Double Diamond execution: Discover → Define → Design → Deliver. D→L→P order enforced.

/evaluate

3-stage verification: Mechanical gates → Semantic check → Human judgment.

🔄

/evolve

Wonder/Reflect/Re-seed loop. Iterates until ontology similarity ≥ 0.95.

👥

Pair Mode v2.1

Navigator-Driver pair programming. 3 plans generated per task. Independent test design.

3-Tier Enforcement. Non-Negotiable.

Auto-detects 30+ frameworks and maps layer rules to your stack. No layer skipping. No reverse deps.

Presentation  ──→  Logic  ──→  Data

P→D direct = ❌ blocked   |   L→P reverse = ❌ blocked   |   DTO at every boundary

See It in Action

Up and Running in 2 Minutes

Option A — Claude Code Plugin

# Simplest way to get started
/plugin marketplace add studioKjm/ai-harness-template
/plugin install harness@studioKjm-harness

Option B — Full Install

# Clone the template
git clone https://github.com/studioKjm/ai-harness-template.git

# Lite — bash only, zero dependencies
./ai-harness-template/init.sh /path/to/your-project

# Pro — Python 3.11+, MCP server + observability
./ai-harness-template/pro/install.sh /path/to/your-project

Start with Lite. It delivers 80% of the value with zero dependencies. MIT Licensed.

Ready to Cage Your AI?

Open source. Framework-agnostic. Drop it into any project.