Skip to content

Latest commit

 

History

10 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

SHP — Smart Human Protocol

Version: 4.0.0
Type: Claude Skill (Enterprise Reasoning Framework)
Format: .skill archive


What is SHP?

SHP is an enterprise-grade reasoning framework for Claude. It governs how Claude thinks — not what it knows.

Rather than adding domain knowledge, SHP installs a structured reasoning discipline:

  • Systematic intent detection and complexity analysis
  • Modular expert coordination across 20+ domains
  • Built-in verification, confidence estimation, and quality gates
  • Consistent memory and project-state tracking
  • Professional, audience-appropriate response delivery

SHP is not a persona. It is not roleplay. It is a reproducible reasoning protocol.


Installation

As a Claude Skill

  1. Upload shp-v3.skill to your Claude session via the paperclip / file attachment UI.
  2. Claude will load the skill and activate SHP reasoning automatically.
  3. Alternatively, prefix your first message with SHP: to activate manually.

Repository Usage

Clone or download this repository. Reference SKILL.md as the skill entry point. All modules are self-contained Markdown files.

shp/
├── SKILL.md           ← Skill entry point
├── core/              ← Reasoning pipeline components
├── modules/           ← Domain expert modules
├── engines/           ← Specialised reasoning engines
├── memory/            ← Project memory system
├── templates/         ← Response templates
├── examples/          ← Worked examples
└── references/        ← External standards and references

Activation

Method Syntax
Prefix trigger SHP: <your request>
Slash command /shp <your request>
Skill upload Upload shp-v3.skill — auto-activates
Session declaration "Use SHP for this session"

Once activated, SHP remains active for the entire session.


Usage Examples

SHP: Design a multi-tenant SaaS authentication system using JWTs and refresh tokens.

SHP: Review this Python function for security vulnerabilities and performance issues.

SHP: Write a technical specification for a real-time event processing pipeline.

SHP: Analyse the trade-offs between microservices and a modular monolith for our use case.

SHP: Generate production-ready Go code for an HTTP middleware chain with rate limiting.

How It Works

Reasoning Pipeline

Every non-trivial request passes through an 18-stage internal pipeline. See core/pipeline.md.

Expert Modules

SHP automatically selects the most relevant expert modules for each request. Modules only activate when relevant — simple questions receive concise answers.

Engines

Specialised engines handle cross-cutting concerns:

Engine Purpose
Confidence Engine Estimates and communicates uncertainty
Verification Engine Reviews drafts before delivery
Multi-Expert Engine Coordinates expert module outputs
Consensus Engine Resolves conflicts between expert recommendations
Threat Model Engine Evaluates security posture
Optimisation Engine Improves response quality and conciseness
Consistency Engine Maintains project-wide coherence
Planning Engine Structures complex multi-step work
Quality Engine Enforces quality gates before delivery

Memory System

SHP tracks architecture decisions, naming conventions, coding standards, and project state across a session. See memory/.


Design Principles

  1. Reason systematically, not superficially. Every response reflects structured analysis.
  2. Calibrate to complexity. Simple tasks get direct answers. Complex tasks get deep reasoning.
  3. Never fabricate certainty. SHP always states confidence level and assumptions explicitly.
  4. Expert depth without expert ego. Modules contribute only when relevant.
  5. Verify before delivering. The first draft is always a draft.
  6. Maintain consistency. Memory tracks decisions so later outputs don't contradict earlier ones.

Extending SHP

See CONTRIBUTING.md for the module development guide.

New modules must define: Purpose, Responsibilities, Activation Criteria, Inputs, Outputs, Dependencies, Examples, Best Practices, Common Mistakes, and Limitations.


Version History

See CHANGELOG.md.


License

See LICENSE.

Releases

Packages

Contributors