Skip to content

PRODUCT.md

Latest commit

 

History

History
52 lines (37 loc) · 4.68 KB

PRODUCT.md

File metadata and controls

52 lines (37 loc) · 4.68 KB
Raw
last_reviewed 2026-02-15

PRODUCT.md

Mission

The doctrine behind the operational layer for coding agents — principles that turn ad-hoc agent usage into bookkeeping, validation, primitives, and flows that compound.

Target Personas

Persona 1: Solo Developer Using AI Agents

  • Goal: Get consistently better results from their agent (Claude Code, Cursor, Codex) without starting from scratch every session
  • Pain point: Inconsistent output quality, context window problems, no bookkeeping across sessions — every session feels like training a new junior developer

Persona 2: Tech Lead Scaling Agent Flows

  • Goal: Coordinate agent-assisted development across a team with shared bookkeeping, validation gates, and multi-agent patterns
  • Pain point: Parallel agent work creates cascading conflicts, learnings from one developer's sessions don't benefit others, no consistent quality bar

Persona 3: Tool Builder / Agent Framework Author

  • Goal: Design agent tooling grounded in proven operational principles rather than inventing patterns from scratch
  • Pain point: No shared vocabulary or reference architecture for agent operations — every framework reinvents the wheel on context management, validation, and bookkeeping persistence

Core Value Propositions

  • Bookkeeping compounds across sessions — Every session makes the next one smarter. Not a wiki nobody reads — extracted learnings become reusable context.
  • Validation becomes an operating rule — The doctrine makes separate review, ratcheting, and evidence-backed closeout part of the normal flow.
  • Primitives and flows scale from one developer to many — Start with one dev + one agent. The same principles extend to multi-agent orchestration.
  • Vendor-neutral doctrine, single-runtime default — Every factor applies whether you run Claude Code, Cursor, Codex, Copilot, or custom setups. The reference implementation defaults to a single runtime so install-to-first-value stays cheap. Cross-vendor coordination is kept as a falsifiable capability — preserved because in real /council --mixed runs the second vendor has changed load-bearing verdicts, not because it sits in the install matrix. The standing proof is auditable.

Competitive Landscape

Alternative Strength Our Differentiation
GSD (Get Shit Done) Spec-driven development with fresh subagent contexts. Fights context rot through task decomposition and phase planning. GSD is a build methodology (spec → plan → implement). We are the doctrine behind the operational layer: bookkeeping, validation, compounding flows, failure harvesting. GSD solves "how to structure one project." We solve "how to operate with agents as a practice."
Superpowers Composable skills framework enforcing TDD, brainstorming, code review, and systematic debugging. Accepted into Claude plugins marketplace. Superpowers is a development methodology (brainstorm → plan → TDD → review). We define the doctrine underneath: context budgets, validation patterns, bookkeeping flywheel, and multi-agent scale patterns. Complementary, not competing.
Agent OS Captures and deploys coding standards across agentic development. Free and open source. Agent OS focuses on coding standards (the rules). We focus on the operational doctrine (the practice). Standards are one input to Factor I; we cover the full lifecycle from research through bookkeeping compounding.
Spec Kitty Spec-driven development with kanban dashboard, git worktrees, auto-merge, acceptance checks. Works across 10+ AI coding agents. Spec Kitty is a project management layer (specs → tasks → kanban → merge). We define how agents should operate while building it — context discipline, validation patterns, bookkeeping extraction, and failure harvesting.

Usage

This file enables product-aware council reviews:

  • /pre-mortem — Automatically includes product perspectives (user-value, adoption-barriers, competitive-position) alongside plan-review judges when this file exists.
  • /vibe — Automatically includes developer-experience perspectives (api-clarity, error-experience, discoverability) alongside code-review judges when this file exists.
  • /council --preset=product — Run product review on demand.
  • /council --preset=developer-experience — Run DX review on demand.

Explicit --preset overrides from the user skip auto-include (user intent takes precedence).