eai-gofer

EAI Gofer

EAI Gofer is a business specification-driven delivery workflow for repositories. It gives teams one shared pipeline from business scenario through validation, keeps the working artifacts in .specify/, and ships across VS Code, Claude Code, Codex, GitHub Copilot, and Gemini.

EAI Gofer is designed to be easy to adopt in an existing repo:

• helps everyone, not just coders, write good code, that delivers a business outcome
• Work with AI to generate what you need whether it is business case, executive summary, technical diagram of otherwise for you and your stakeholders to know what will be built, not find out it is wrong later
• one core 0-6 delivery pipeline
• closed-loop goal reconciliation with repo-owned goal ledgers and drift checks
• repo-owned artifacts and templates
• install paths for VS Code and AI coding CLIs
• generated command surfaces that stay aligned across hosts

Quick Start

1. Install the VS Code extension or add the public plugin marketplace for your preferred CLI.
2. For a first EAI Platform app, run /gofer:eai-first-run. It checks Git, Node.js, npm, EAI CLI, login, tenant access, the EAI app template, and the Gofer repo scaffold before the pipeline starts.
3. If you only need to add Gofer to an existing repo, run Gofer: Initialize Repository in VS Code or /gofer:bootstrap-workspace from a CLI host.
4. Start with /0_business_scenario and move through the core pipeline.

If /0_business_scenario is unknown in a new repo, the host has not loaded the Gofer plugin or repo commands yet. Install/update the plugin first, then run /gofer:eai-first-run to prepare the workspace and start the app build.

For copy-paste commands across VS Code, Claude Code, Codex, Copilot, and Gemini, see the 5-minute first run guide.

Core Pipeline

Stage	Command	Main output
Business scenario	/0_business_scenario	Full pipeline kickoff
Research	/1_gofer_research	research.md
Specify	/2_gofer_specify	spec.md
Plan	/3_gofer_plan	plan.md, data-model.md, contracts/
Tasks	/4_gofer_tasks	tasks.md, traceability.md, issues.md
Implement	/5_gofer_implement	Code and documentation changes
Validate	/6_gofer_validate	Validation artifacts and final review evidence

/6_gofer_validate is the terminal quality gate. The previous standalone engineering-review stage is now folded into validation.

Gofer now treats the pipeline as a closed loop, not just a straight line:

• /1_gofer_research seeds goal-ledger.json with business goals, metrics, owners, delivery states, and re-loop triggers.
• /4_gofer_tasks keeps traceability.md as the requirement-to-task-to-code contract.
• /6_gofer_validate runs the objective outcome gate, refreshes code/test evidence, and writes goal-rebaseline-report.md when goals, assumptions, contracts, UX scope, or implementation drift.

Optional helpers stay outside the core 0-6 flow:

• /0a_problem_validation
• /7_gofer_save
• /8_gofer_resume
• /9_gofer_tests
• /7a_stakeholder_comms
• /gofer:check-workspace
• /gofer:bootstrap-workspace
• /gofer:eai-first-run

Model And Cost Policy

EAI Gofer bootstraps a repo-owned model policy at:

.specify/memory/gofer-model-policy.yaml

The shipped default comes from .specify/templates/gofer-model-policy.yaml. Bootstrap creates the memory copy when it is missing and does not overwrite local edits.

Default posture:

• Claude: Haiku for simple scouting, Sonnet for normal work, Opus for hard security/architecture/release gates.
• Codex/OpenAI: GPT mini for simple coding, GPT nano only for mechanical locate/classify/summarize work, GPT-5.3-Codex or flagship GPT for hard tool-heavy coding and arbitration.
• Gemini: Flash-Lite for cheap large-context scanning, Flash for normal synthesis, Pro for hard large-context architecture/research gates.
• Copilot: Auto for simple/default work; ask before selecting a paid/high-tier picker model for hard review.

Install

VS Code

Recommended: install from the VS Code Marketplace so users receive normal Marketplace updates. Manual .vsix installs remain supported, but VS Code does not auto-update VSIX installs by default.