Skip to content

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

1 Commit
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Rivulet Analytics operating wiki

This is the shared operating reference for Rivulet employees and trusted agents. It connects product, data, engineering, delivery, security, and agent work so that a second person can understand what is current, who owns a decision, what has actually happened, and how work should continue.

Rivulet is maintained as an open reference implementation. Its design rationale, adaptation guidance, and provenance live in the separate agent-wiki-reference repository.

Reuse and contribution

This repository is available under the MIT License. Adapt the structure to your own organization; do not copy Rivulet's names, routes, or operating assumptions as if they were your own facts. Read CONTRIBUTING.md before proposing a change and SECURITY.md before reporting a potential security concern.

Start here

If you need to… Read
Orient to Rivulet and current priorities Company and market and portfolio priorities.
Begin work as an employee or agent AGENTS.md, context layering, and employee onboarding.
Start or continue an Estuary task Project Estuary, its project instructions, and memory index.
Handle a data/model/release question Source contracts, model card, and runtime/release.
Work with an agent capability Agent roster, task state, and agent-lab.
Change a system, source, access, or external message Account routing, evidence and external actions, and the relevant system card.

The documentation map links every maintained page.

Operating principles

Principle Working meaning
Name authority A task identifies who may decide, approve, execute, and verify.
Keep truth systems distinct A board item, meeting packet, PR, deployment, customer message, and observed outcome are different facts.
Preserve evidence posture Source type, authority, freshness, limitations, and correction path travel with a claim.
Keep actions reviewable Agents prepare, compare, validate, and hand off; named humans approve external or irreversible actions.
Retain useful failure A correction, incident, or stop condition improves the next task rather than disappearing into chat history.
Keep context layered Shared wiki, project memory, task state, employee-local configuration, and native-system truth have different owners.

State language

State Meaning
prepared A reviewable artifact exists.
approved The named owner accepted a bounded next action.
executed The authorized action occurred in the relevant system.
verified The specified readback supports the stated result.
blocked / needs assignment Authority, source, access, or other required condition is absent.
stale / superseded Later information changed how the earlier record should be used.

Before changing a page or system, read AGENTS.md.

Releases

Packages

Contributors

Languages