By Hypho
Ledger — auditable AI-assisted development protocol. Skills-based agent guidance with file-based state, behavior contracts, runtime evidence, and shell-enforced guards.
Use when phase=contract and contract exists. Implements features using TDD against the behavior contract. Enforces RED-GREEN-REFACTOR for every FC entry, component-level checks, and architecture spine compliance.
Use when phase=pid and PID Card exists. Generates behavior contract from PID Card with FC/NF entries, scope boundaries, and test coverage mapping.
Use when a new feature is requested and no active unshipped feature exists. Defines feature intent, maps to product spine, detects risk boundaries, and generates PID Card. This is constrained intent capture, not open brainstorming.
Use every 3-5 shipped features or when intent drift is suspected. Reviews product drift, feature stacking, architecture violations, and extracts reusable patterns.
Use before the first feature or when product direction changes. Assesses Ledger applicability, identifies risk boundaries, and optionally generates feature dependency graph.
Executes bash commands
Hook triggers when Bash tool is used
Modifies files
Hook triggers on file write and edit operations
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
A lightweight protocol framework for auditable human-AI software development | v0.2.0 中文: README.zh.md
Ledger is a lightweight protocol framework for building software with AI while keeping product intent, implementation scope, and verification evidence explicit.
It turns AI-assisted development from an open-ended chat into a staged workflow: define intent, write a behavior contract, implement against that contract, verify with real outputs, then archive what changed.
Ledger keeps the daily feature loop lightweight, but feature work is anchored by two global spines: the Product Spine in PAD.md and the Architecture Spine in architecture.md. A feature should explain where it fits in the core business flow and which module / entity boundaries it touches before it enters build.
Ledger is designed for product-minded builders, solo developers, and small teams who want AI to move faster without losing control of scope, state, and quality.
It is not a code generator, an agent scheduler, or a replacement for CI/CD. It is the protocol layer that keeps human decisions and AI execution aligned.
Core idea: Define behavior before implementation. Verify behavior before shipping.
Ledger v0.2.x uses a skills-based architecture. Skills are composable protocol documents that guide the agent through each phase. They auto-trigger based on project state.
| Skill | Phase | Purpose |
|---|---|---|
using-ledger | Session start | State awareness, phase routing, guard enforcement |
ledger-scope | Pre-feature | Ledger applicability assessment |
ledger-pid | pid | Intent definition + boundary detection |
ledger-contract | contract | Behavior contract generation |
ledger-build | build | TDD implementation with complexity scoring |
ledger-verify | verify | Subagent-driven adversarial verification |
ledger-ship | ship | Test suite, acceptance, archive |
ledger-retro | Every 3-5 features | Protocol quality review |
How it works:
.ledger/ in the project and injects using-ledger into the agent contextstate.md to determine the current phaseIn Claude Code, skills are provided via a plugin.
Use Ledger when:
Avoid Ledger when:
Before adopting, check whether your project falls within Ledger's applicable scope.
/ledger.buildLedger will actively stop you in these situations but does not propose solutions. Pair with specialized reviews (security / performance / DBA). The detect-and-halt behavior is itself one of the framework's deliverables.
Ledger does not engage with these — use other toolchains.
Install details: INSTALL.md Full usage guide: USAGE.md Global Spine Lite guide: docs/concepts/global-spine-lite.md Design Attachments Lite guide: docs/concepts/design-attachments-lite.md Existing project adoption: docs/migration/adopt-global-spine-lite.md
# Recommended: install directly into a project from GitHub
curl -fsSL https://raw.githubusercontent.com/Hypho/ledger-protocol/master/scripts/install-from-github.sh | bash -s -- --target your-project --mode auto
npx claudepluginhub hypho/ledger-protocol --plugin ledgerFeature development with code-architect/explorer/reviewer agents, CLAUDE.md audit and session learnings, and Agent Skills creation with eval benchmarking from Anthropic.
Production-grade engineering skills for AI coding agents — covering the full software development lifecycle from spec to ship.
Permanent coding companion for Claude Code — survives any update. MCP-based terminal pet with ASCII art, stats, reactions, and personality.
Claude harness - A harness for solo developers (Vibecoders) to handle full-cycle contract development.
Complete Claude Code configs for mobile development - Android, iOS, and Kotlin Multiplatform
Core skills library for Claude Code: TDD, debugging, collaboration patterns, and proven techniques