REI HARNESS / PRODUCT DOSSIER

SHEET 01 / 06

v0.3.0 · EN

Product · Spec Driven Development

What REI is

What REI Harness is, what problem it solves and why the process matters before the commands.

Source: README.md, docs/product-strategy.md and .rei/docs/

P.01Problem

The problem: deciding, planning and executing in the same conversation

When deciding what to build, planning how and making the change all happen in the same conversation, nothing is recorded and nobody can review the decision. AI speeds up writing, but it does not guarantee that the right thing is being built.

  1. 01

    Building blind

    Code is written before the scope is clear.

  2. 02

    Hallucinations

    Invented APIs or data presented as fact.

  3. 03

    Unauthorized decisions

    Architecture or scope changed without approval.

  4. 04

    Out-of-bounds actions

    Environments and data that should not be touched.

  5. 05

    Rework

    Code that is fast to write and expensive to fix.

P.02Mechanism

Featured sheet

The proposal: specification, human approval and evidence

REI Harness is a Spec Driven Development harness that installs in any repository and coordinates AI agents behind a mandatory human approval gate. It separates deciding, planning and executing, and leaves the trail in the repository itself.

  1. 01

    Specification first

    The plan exists before the first line of code.

  2. 02

    Planning approval

    No Work Item moves to implementation without an explicit sign-off; silence is not approval.

  3. 03

    Separate roles

    Leader, Spec Author, Implementer and Reviewer each have their own contract; the Initializer customizes the project documentation.

  4. 04

    Guardrails

    Defined scope, mandatory verification and hard rules.

  5. 05

    Traceability

    Planning, progress, decisions and history live in .rei/.

  6. 06

    Living documentation

    What the site states can be traced to the repository or to a cited source.

  7. 07

    Runtime-agnostic and local-first

    The role contract and the CLI are the value; the adapter is optional.

P.03Evidence

External evidence

Most developers use AI, and many distrust how accurate it is. This site does not make up its own figures: each case links to its source and sticks to what the study shows.

Cited external evidence · each case links to its source

  1. An agent deleted a production database during a code freeze, ran destructive commands without authorization and generated fake data.

    Implication

    The clearest case of why human approval, separate environments and limits on actions are needed.

  2. 02

    METR

    2025

    Randomized study with experienced open source developers: with AI they took 19% longer, although they believed they had been 20% faster.

    Implication

    Perceived productivity is not evidence. You have to measure and verify.

  3. 03

    GitClear

    2025–2026

    More duplication, more copy/paste, more churn and less refactoring as AI authorship grows.

    Implication

    Speed without design degrades maintainability.

  4. 04

    Veracode

    2025–2026

    About 45% of code generation tasks introduced a known vulnerability; security does not necessarily improve with larger models.

    Implication

    Review and guardrails are not optional.

  5. AI adoption is associated with higher throughput and lower delivery stability: AI amplifies a system's strengths and weaknesses.

    Implication

    The existing process decides the outcome, not the tool.

  6. 84% use or plan to use AI; 46% distrust the accuracy of its results and only 3% highly trust them.

    Implication

    There is demand for verifiable trust, not just for speed.

  7. On a bounded task, the group with Copilot finished 55.8% faster.

    Implication

    AI does speed up specific tasks; the problem is the lack of process.

P.04Principles

Product principles

  1. 01

    The why before the what

    The process and the problem justify the commands.

  2. 02

    Human approval is the product

    The flow makes the gate visible; it does not hide it.

  3. 03

    Documentation is the source of truth

    Every claim can be traced.

  4. 04

    Showing limits builds trust

    The viewer is read-only, GitHub Issues is optional and direct mode has restrictions.

  5. 05

    One source per fact and minimal dependencies

    Commands, states and roles are defined once.

P.05Limits

What REI is not

  1. Not a code generator

    Nor an autonomous agent that works on its own.

  2. Not the CLI repository

    The code lives in its own repository and this site documents it.

  3. Not the spec viewer

    rei view is local and read-only.

  4. Not a control panel

    The site does not run commands or modify Work Items.

  5. Not a metrics showcase

    No customers, testimonials or figures that cannot be shown.

P.06Audiences

Who it is for

P.07Strengths

Distinguishing strengths

  1. SDD from minute zero

    Code does not start until an approved plan exists.

  2. Mandatory human approval

    No Work Item moves to implementation without an explicit decision; silence is not approval.

  3. Explicit, separate roles

    Each role has a contract and does not step into another's work.

  4. Guardrails

    Defined scope, mandatory verification, diff-based review and hard rules.

  5. Traceability

    Planning, progress and decisions stay in the repository, not in the chat.

  6. Living documentation

    The source of truth is the repository and it is updated with the work.

  7. Runtime-agnostic

    The value is in the role contract and the CLI; the adapter is optional.

  8. Optional GitHub Issues

    Issues, branches and PRs are a projection of the work, not the source of truth.

  9. Single binary

    One executable with no runtime dependencies.

  10. Local-first

    Everything happens on the user's machine and repository; the CLI does not call external services.

  11. Minimal dependencies

    The platform and own code before third-party libraries.