probe

Probe

A multi-spectrum research context architecture for deep technical investigation.

---

What is Probe?

Probe is a Claude Code plugin that conducts systematic, multi-dimensional technical research following the Gradient architectural pattern.

Like a space probe uses multiple instruments to analyze targets from multiple spectra, this system uses multiple research dimensions to investigate technical topics comprehensively.

Give it a term or query, and Probe investigates from 10 distinct perspectives:

• Fundamentals - What it is, why it exists
• Architecture - How it works internally
• Technical Deep-Dive - Advanced techniques, algorithms
• Academic Research - Papers, theoretical foundation
• Integration - How to adopt, prerequisites
• Performance - Benchmarks (realistic, not marketing)
• Use Cases - When to apply
• Applicability - When to use AND when NOT to use
• Implementation - Code examples, production patterns
• Critical FAQ - Honest Q&A, including "NO" answers

Probe reproduces structured research methodology through layered context (SPEC → CONTEXT → PROMPTS), ensuring consistent, thorough investigations that go beyond naive marketing acceptance.

---

Why "Probe"?

Semantic Rationale

Probe = The investigative agent, the system that explores and discovers.

The name draws inspiration from space probes (Voyager, Mars rovers, etc.):

PROBE (Voyager, Cassini, etc.)
├─ Spectrometer   → UV spectrum
├─ Camera         → Visible spectrum
├─ IR Sensor      → Infrared spectrum
└─ Radar          → Radio spectrum

Just as space probes use multiple instruments to analyze targets from multiple spectra, this system uses multiple research dimensions to understand technical topics comprehensively.

Probe is the WHAT (the investigative system). Spectrum is the HOW (multi-dimensional analysis method).

---

Philosophy

"Like a space probe scanning distant worlds with multiple instruments, Probe examines technical topics through multiple analytical lenses."

Research isn't one-dimensional. Every query deserves investigation from libraries, techniques, comparisons, real-world applications, AND critical assessment of limitations.

Key principles:

• Beyond naive acceptance - Challenge claims, verify independently
• Limitations as prominent as benefits - Honest assessment
• Multi-spectrum coverage - All 10 dimensions investigated
• Context-specific applicability - Does it fit YOUR use case?
• Critical FAQ - Honest "NO" answers when appropriate

---

Quick Start

Installation

Install as Claude Code plugin (assumes Gradient structure for plugins).

Usage

/probe:research <topic>

Examples:

/probe:research lmcache
/probe:research llm-cache
/probe:research vector-databases

What happens:

1. Creates ~/work/sources/probes/probe-<topic>/
2. Creates private GitHub repo research-<topic>
3. Investigates all 10 research spectra
4. Generates comprehensive research artifacts
5. Commits and pushes to GitHub
6. Returns summary with key findings and recommendation

Time: 10-30 minutes autonomous execution (no pausing mid-research)

---

Architecture

Probe follows the Gradient pattern:

probe/
├── spec/                           # SPECS - Normative (WHAT to research)
│   ├── research-methodology-spec.md      # Deep vs naive research
│   ├── research-spectra-spec.md          # 10 investigation dimensions
│   ├── critical-questions-spec.md        # Mandatory questions per spectrum
│   └── validation-spec.md                # Completeness/depth/quality criteria
│
├── context/                        # CONTEXT - Applied (HOW to research)
│   ├── research-patterns.md              # Investigation patterns by topic type
│   ├── anti-naive-guide.md               # Avoiding superficial research
│   ├── depth-calibration.md              # Adaptive depth based on context
│   ├── critical-thinking-guide.md        # 10-stage analysis framework
│   └── examples.md                       # LMCache research analyzed
│
├── prompts/                        # PROMPTS - Orchestration
│   ├── load-research-context.md          # Load specs + context (SSOT)
│   ├── conduct-research-workflow.md      # Main research workflow
│   └── setup-research-environment-workflow.md  # Repo creation workflow
│
└── scripts/                        # SCRIPTS - Automation
    ├── generate-repo-name.sh             # Normalize topic to repo name
    └── create-research-repo.sh           # Create local + GitHub repo

agents/                             # AGENTS - Autonomous execution
└── research-conductor.md                 # Multi-spectrum research agent

commands/                           # COMMANDS - Entry point
└── research.md                           # /probe:research command