Federated Wiki serves a dual role in the Hitchhiker’s Project: 1. As a **presentation layer** for agent-generated mini-apps and artifacts 2. As a **shared repository of structured prompts and agent plans**
This dual use allows for a seamless loop between Agent Action and Human Review, where the same system used to display outputs can also direct future inputs.
This strategy transforms Federated Wiki into a living repository of both **actionable plans** and **presentable results**, closing the loop between intent and outcome.
# Vibe-Coding Presentation Layer Federated Wiki supports embedding HTML and JavaScript using the `frame` plugin. This allows agents to: - Generate visual mini‑apps or Vibelets - Save these directly into the wiki's `assets/` folder - Instantly view and interact with them via a wiki page
This makes Federated Wiki ideal for vibe coding workflows: agents generate live demos, small tools, or prototypes that are viewable and editable in the browser, without the need for separate hosting - github.com 
# Agent Prompts and Plan Repository Each wiki page is stored as a structured JSON file with fields like `title`, `story[]`, and `journal[]`. This structure can be used to store: - Agent Instructions - Multi-step Workflows - Memory Checkpoints - Tool References or required resources These plans can be fetched directly via HTTP calls, whether hosted locally or remotely:
GET http://localhost:3000/wiki/page/my-plan.json
Agents (e.g. Claude Code, CrewAI, Open Interpreter) can use these URLs as input contexts, treating wiki pages as Living Prompts.
- The Plan
Agents can fetch these plans, interpret them step-by-step, and update the wiki with results or notes.
# Why This Is Novel - **Pages are instantly accessible**: Agents can pull prompts via standard HTTP, no API layer required - **Federated architecture**: Anyone can fork or remix a plan by editing the page locally or federating changes - **Editable by humans and agents**: Prompts evolve through human refinement, agent feedback, or review - **JSON-native**: Because wiki pages are stored in JSON, they’re ideal for structured agent use - **Composability**: A single plan can link to other plans, wiki pages, or assets, allowing agents to build complex flows from modular documents
# Practical Applications - Write a plan once, reuse it across machines and agent systems - Create a gallery of plans others can fork and improve - Store "agent recipes" that describe common tasks (e.g. write docs, build a quiz, refactor code) - Log revisions and branches using wiki’s federated history and journaling
# Next Steps 1. Draft a set of core plans and publish them to your local wiki 2. Create a "Plan" template page with clear sections (goal, steps, output, etc.) 3. Teach agents to fetch, interpret, and execute these plans step by step 4. Use LiveCode to list available plans and send selected ones to your agents 5. Allow agents to write their own plan pages as output (agent-generated plans!)