Skip to content

jcalyon/wakefile

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

3 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

wakefile

A structured methodology for human/agent collaborative projects. Wakefile gives AI agents and their human operators a shared operational layer: loop management, role definitions, content production, and progressive elaboration across long-running work.

Quick start

  1. Clone or copy this template into your project root.
  2. Place any background material (project briefs, prior art, domain context) into seeds/.
  3. Open a conversation with your agent and run ::init.
  4. The init conversation configures your project identity, domain, and initial cycle. It produces a populated WAKEFILE.md, a DOMAIN.md configuration record, and an INTAKE.md handoff for the planner agent.
  5. From there, open your first cycle with ::open-cycle and begin work.

What is in this template

WAKEFILE.md is the root entry point. It contains the project identity and a directory guide for .wake/.

.wake/ is the operational directory. It contains agent role definitions, command references, templates for cycles and contracts, content production scaffolds, and ritual specifications. All of these are pre-populated as a scaffold; ::init and subsequent commands fill them with project-specific content.

seeds/ is an optional directory for pre-initialization material. Place anything here that should inform the ::init conversation.

Full methodology documentation, including loop definitions, contract lifecycle, beacon usage, and migration paths, is available at wakefile.com. The .wake/COMMANDS.md file is the authoritative in-repo command reference.

Core concepts

Loops. Work is organized in nested loops: era (optional macro container), cycle (primary recurrence unit), phase (major section within a cycle), and step (atomic unit of work). Each loop level has open and close commands with associated rituals.

Agents. Wakefile defines four agent roles: planner, generator, evaluator, and reporter. A single human or AI can fill multiple roles. The roles define responsibilities, not identities.

Commands. All operations use the :: trigger prefix. Commands open and close loops, manage state, produce content, and handle escalation. See .wake/COMMANDS.md for the full reference.

Content. Three artifact types accumulate over a project's life: eductions (reasoning traces extracted during work), dispatches (narrative accounts at loop boundaries), and claims (durable assertions that mature through validation).

Configuration

.wake/ is gitignored by default. This means your operational state stays local. To version-control your wake files (recommended for team projects or long-running work), remove .wake/ from .gitignore.

Documentation

Full documentation is available at wakefile.com.

License

Apache License 2.0. See LICENSE.md.

About

A structured methodology for human/agent collaborative projects.

Resources

License

Stars

Watchers

Forks

Packages

 
 
 

Contributors