SPECTRE

Product Introduction

1. Definition: SPECTRE is an agentic coding workflow plugin for Claude Code, designed as a structured slash-command system (/Scope, /Plan, /Execute, /Clean, /Test, /Rebase, /Evaluate) to manage AI-assisted software development. It falls under the technical category of AI-powered development automation tools.
2. Core Value Proposition: SPECTRE exists to eliminate ambiguity in AI coding workflows, enabling developers to generate high-quality, production-ready code consistently. Its primary value lies in transforming ad-hoc AI interactions into a repeatable, canonical-document-driven process that scales from small fixes to large features across any codebase.

Main Features

1. Canonical Documentation Generation:
   • How it works: SPECTRE prompts systematically create structured markdown documents (scope.md, ux.md, plan.md, tasks.md) stored in docs/tasks/{topic}/specs. These docs define requirements, user flows, technical designs, and tasks upfront.
   • Technology: Uses Claude Code's reasoning to interrogate user intent, research codebases, and output standardized specs, enforcing rapid waterfall principles for AI development.
2. Session Memory & Handoff (/spectre:handoff):
   • How it works: Generates session status reports stored in docs/tasks/{topic}/session_logs. When starting a new session, the @spectre:sync subagent reviews the last 3 reports, merging them into a continuous context.
   • Technology: Implements a trailing context snapshot system using Claude's file handling and subagent orchestration to overcome context window limits, enabling multi-session feature development.
3. Evaluate & Learn (/spectre:evaluate):
   • How it works: Conducts an automated architecture review using Opus 4.6 subagents and captures durable knowledge (gotchas, decisions, patterns) into .claude/skills/{feature_name}/skill.md files.
   • Technology: Employs a Hook + Skill Loop; a SessionStart hook injects relevant skills into context, and skills auto-load based on trigger words (e.g., mentioning "auth" loads feature-auth-flows skill), creating compounding institutional memory.
4. Subagent Orchestration:
   • How it works: Dispatches specialized AI agents (e.g., @spectre:dev for MVP coding, @spectre:tester for risk-based testing, @spectre:web-research for deep dives) during commands like /execute or /test.
   • Technology: Leverages Claude Code's multi-agent collaboration framework to parallelize tasks (e.g., simultaneous coding, review, validation) within a single context window.

Problems Solved

1. Pain Point: AI coding ambiguity leading to inconsistent output, spaghetti code, and context collapse during complex or multi-session development. SPECTRE enforces spec-driven development via canonical docs.
2. Target Audience:
   • Full-Stack Developers building features solo or in small teams.
   • Technical Product Managers owning feature definition and delivery.
   • Engineering Leads standardizing AI-assisted workflows across teams.
3. Use Cases:
   • Building complex features in large (>100k LOC) codebases (React Native, Tauri/Rust/React) with reliable outcomes.
   • Maintaining context during long-running development tasks spanning multiple Claude sessions.
   • Onboarding AI agents onto existing projects by capturing and auto-loading codebase knowledge.

Unique Advantages

1. Differentiation: Unlike vanilla Claude Code or fragmented prompt libraries, SPECTRE provides an integrated, end-to-end workflow with built-in context persistence, knowledge capture, and quality gates (clean/test/rebase). Competitors lack its canonical doc enforcement and session memory system.
2. Key Innovation: The Hook + Skill Loop for autonomous knowledge retrieval. This transforms past work into actionable context, allowing Claude to "remember" project specifics without manual prompting or redundant research – a breakthrough in AI agent memory.

Frequently Asked Questions (FAQ)

1. How do I install SPECTRE in Claude Code?
   Run /plugin marketplace add Codename-Inc/spectre followed by /plugin install spectre@codename within Claude Code to install this AI coding workflow plugin.
2. Can SPECTRE handle large legacy codebases?
   Yes, SPECTRE's canonical docs and session memory are explicitly designed for large codebases (tested on 250k+ LOC), using subagents like @spectre:analyst and @spectre:patterns to navigate and document complex systems.
3. Does SPECTRE replace traditional testing or CI/CD?
   No, SPECTRE's /spectre:test command generates risk-aware tests (unit/integration) based on behavior, complementing existing pipelines. /spectre:rebase prepares code for merge but relies on external CI/CD for final checks.
4. How does SPECTRE reduce AI "hallucination" in coding?
   By enforcing ambiguity elimination via rigorous scoping (/scope) and planning (/plan), and using validated canonical docs as the source of truth for execution, minimizing guesswork by AI agents.
5. Is SPECTRE suitable for non-UI backend development?
   Absolutely. While /spectre:ux_spec targets UI flows, core commands like /scope, /plan, /execute, and /test are backend-agnostic and used for API, database, or infrastructure tasks.