RepoAIReady 0.1.0-preview.10

RepoAIReady

RepoAIReady is a .NET command-line tool that evaluates GitHub repositories for AI coding-agent readiness. It collects repository evidence through the GitHub API, judges it against a Markdown rubric, and writes console, Markdown, and JSON reports.

Features

• Evaluates one or more GitHub repositories for AI coding-agent readiness using a Markdown judge rubric.
• Scores repositories across documentation, style and validation, testing, build infrastructure, and AI context.
• Collects targeted GitHub evidence from README files, docs, workflows, dependency files, tests, source structure, dev containers, and Copilot-specific guidance.
• Supports GitHub Copilot, OpenAI-compatible, and deterministic offline judging backends.
• Runs repository evaluations in parallel with configurable worker count and live progress reporting.
• Produces console dashboards plus Markdown and JSON reports, including per-repository details and aggregate summaries.
• Highlights top strengths, highest-impact improvements, and uncertainties for each repository.
• Supports .env configuration, explicit tokens, model overrides, minimum-score exit codes, and private-repository evidence collection.

Prerequisites

• .NET SDK installed so you can install and run the tool with dotnet tool.
• Access to a judge backend:
  • Default copilot backend: a logged-in GitHub Copilot CLI/SDK account, or COPILOT_TOKEN/GITHUB_COPILOT_TOKEN set; the packaged tool currently supports this backend only on Windows x64.
  • openai backend: OPENAI_API_KEY set, with OPENAI_ENDPOINT if you use an OpenAI-compatible service such as Azure OpenAI.
  • deterministic backend: no LLM credentials required; useful for offline smoke tests and CI.
• Optional GITHUB_TOKEN or GH_TOKEN for collecting repository evidence from private repositories or to avoid unauthenticated GitHub API rate limits.

Installation

Install the latest preview package from NuGet.org:

dotnet tool install --global RepoAIReady --prerelease

If RepoAIReady is already installed, update it with:

dotnet tool update --global RepoAIReady --prerelease

Then run it with the repo-ai-ready command:

repo-ai-ready microsoft/vscode

Until the package is published, install from a local package build:

dotnet pack src\RepoAIReady\RepoAIReady.csproj --configuration Release
dotnet tool install RepoAIReady --global --add-source artifacts\nupkg

Usage

repo-ai-ready <org/repo> [org/repo ...] [--judge-file <path>] [--output <dir>] [--format console|markdown|json|all] [--backend copilot|openai|deterministic] [--env-file <path>] [--github-token <token>] [--copilot-token <token>] [--openai-key <key>] [--openai-endpoint <uri>] [--model <model>] [--min-score <0-100>]

RepoAIReady uses its bundled ai-readiness-llm-judge.md by default. Use --judge-file <path> to load a custom rubric.

Examples:

repo-ai-ready microsoft/vscode
repo-ai-ready microsoft/vscode dotnet/runtime --format all --output reports
repo-ai-ready microsoft/vscode --judge-file custom-judge.md --backend deterministic

Authentication and backends

RepoAIReady loads .env from the current directory by default. Copy .env.example and fill in the values you need:

Copy-Item .env.example .env

Supported environment variables:

Backends:

Copilot packaged-tool platform support

The NuGet package currently bundles the GitHub Copilot CLI native binary only for Windows x64 (runtimes/win-x64/native/copilot.exe). The default copilot backend is therefore supported by the packaged tool only on Windows x64; on Linux, macOS, or Windows ARM64, use --backend deterministic for offline heuristic judging or --backend openai for OpenAI-compatible judging. Unsupported Copilot platforms fail with an explicit error instead of silently implying cross-platform Copilot startup.

During dotnet pack, the project downloads @github/copilot-win32-x64 from registry.npmjs.org using the CopilotCliVersion supplied by the GitHub Copilot SDK package. The current pack target does not pin an additional SHA-256 checksum, so pack and release should run only in trusted CI; the workflows verify the expected package entries and run an installed-tool deterministic smoke evaluation before publishing.

Development

Build and test:

dotnet restore RepoAIReady.sln
dotnet build RepoAIReady.sln --configuration Release --no-restore
dotnet test RepoAIReady.sln --configuration Release --no-build

Package and smoke-test the tool:

dotnet pack src\RepoAIReady\RepoAIReady.csproj --configuration Release --output artifacts\nupkg
$toolPath = Join-Path (Resolve-Path artifacts).Path "tool-smoke"
$smokeOutput = Join-Path (Resolve-Path artifacts).Path "smoke-output"
Remove-Item -Recurse -Force $toolPath, $smokeOutput -ErrorAction SilentlyContinue
dotnet tool install RepoAIReady --tool-path $toolPath --add-source artifacts\nupkg --no-cache
& (Join-Path $toolPath "repo-ai-ready") lucabol/repo-ai-ready --backend deterministic --format json --output $smokeOutput --parallelism 1
Get-ChildItem $smokeOutput -Recurse -Filter aggregate-report.json

Releases

Releases are driven by version tags:

git tag v0.1.0
git push origin v0.1.0

The release workflow builds, tests, packs, publishes to NuGet.org with Trusted Publishing, and creates or updates a GitHub Release.

Configure the trusted publisher on nuget.org with these values:

Trusted Publishing uses OIDC, so the workflow does not require a NuGet API key secret.

License

This project is licensed under the MIT License. See LICENSE.