Skip to content

agentrc.workflow.md

Latest commit

 

History

History
216 lines (147 loc) · 6.83 KB

agentrc.workflow.md

File metadata and controls

216 lines (147 loc) · 6.83 KB

I'll help you understand the workflow for running through all the commands in this AgentRC project. Let me first load the repository structure and command reference to give you a comprehensive guide.Let me try a different approach to find the commands:Based on the semantic and lexical search results, here's a comprehensive workflow for running through all the commands in the AgentRC project:

Complete AgentRC Command Workflow

The project has 9 main commands. Here's the recommended execution order for a complete walkthrough:

1. agentrc init — Interactive Setup

npx agentrc init [path]

Purpose: Guided interactive setup for a repository

  • Analyzes the repo
  • Generates instructions and configs
  • Optionally clones from GitHub or Azure DevOps
  • Use --yes to skip prompts, --github or --provider azure to specify source
  • Creates .agentrc-cache for remote repos

2. agentrc analyze — Inspect Repository

npx agentrc analyze [path]
npx agentrc analyze --output report.json    # Save report
npx agentrc analyze --output report.md      # Markdown format

Purpose: Detect languages, frameworks, monorepo structure, areas

  • Scans for: languages, package manager, build system, test setup
  • Detects monorepo workspaces
  • Output: JSON or Markdown

3. agentrc readiness — AI Readiness Assessment

npx agentrc readiness [path]
npx agentrc readiness --visual              # HTML report
npx agentrc readiness --output report.html
npx agentrc readiness --per-area            # Per-area breakdown
npx agentrc readiness --fail-level 3        # CI gate (fail if < level 3)

Purpose: Score repo across 9 maturity pillars:

  • Repo Health: Style, Build, Testing, Docs, Dev Environment, Code Quality, Observability, Security
  • AI Setup: AI Tooling setup

Maturity levels: 1 (Functional) → 5 (Autonomous)

4. agentrc instructions — Generate Copilot Instructions

npx agentrc instructions                    # Default: .github/copilot-instructions.md
npx agentrc instructions --areas            # Root + all detected areas
npx agentrc instructions --areas-only       # File-based area instructions only
npx agentrc instructions --area frontend    # Single area
npx agentrc instructions --strategy nested  # AGENTS.md + detail files
npx agentrc instructions --claude-md        # Generate CLAUDE.md (nested)

Purpose: Generate AI instructions for Copilot SDK

  • Strategies: flat (single file) or nested (hub + area details)
  • Integrates with Copilot CLI for model selection

5. agentrc eval — Evaluate Instructions

npx agentrc eval --init                     # Scaffold eval config (agentrc.eval.json)
npx agentrc eval --init --count 10          # Generate 10 test cases
npx agentrc eval agentrc.eval.json          # Run evaluation
npx agentrc eval --model gpt-4 --judge-model claude-sonnet-4.5
npx agentrc eval --fail-level 80            # CI gate: fail if pass rate < 80%
npx agentrc eval --output results.json      # Save results

Purpose: Measure instruction quality by comparing with/without instructions

  • Judge model rates responses
  • Generates trajectory viewer

6. agentrc generate — Generate Configs

npx agentrc generate instructions [path]    # Same as `agentrc instructions`
npx agentrc generate agents [path]          # Generate AGENTS.md
npx agentrc generate mcp [path]             # MCP server config
npx agentrc generate vscode [path]          # VS Code settings

Purpose: Generate ecosystem configs

  • Supports monorepos with --per-app

7. agentrc pr — Create PR with Configs

npx agentrc pr owner/repo                   # GitHub
npx agentrc pr org/project/repo --provider azure
npx agentrc pr owner/repo --branch custom-branch

Requires:

  • GITHUB_TOKEN / GH_TOKEN for GitHub
  • AZURE_DEVOPS_PAT for Azure DevOps

Purpose: Clone repo, generate instructions + configs, create PR

8. agentrc batch — Batch Process Multiple Repos

npx agentrc batch owner/repo1 owner/repo2 owner/repo3
npx agentrc batch --provider azure org/project/repo1 org/project/repo2
npx agentrc batch --output results.json

Purpose: Run instructions generation across multiple repos in parallel

9. agentrc batch-readiness — Batch Readiness Report

npx agentrc batch-readiness
npx agentrc batch-readiness --output report.html
npx agentrc batch-readiness --policy ./policies.json

Purpose: Generate HTML report comparing readiness across multiple repos

10. agentrc tui — Interactive Terminal UI

npx agentrc tui                             # Full interactive mode
npx agentrc tui --repo /path/to/repo
npx agentrc tui --no-animation              # Skip banner

Purpose: Interactive menu for generation, evaluation, batch workflows

  • Default when run with no args

Recommended Full Workflow

# 1. Set up repo
npx agentrc init /path/to/repo

# 2. Analyze structure
npx agentrc analyze /path/to/repo --output analysis.json

# 3. Assess readiness
npx agentrc readiness /path/to/repo --visual --output readiness.html

# 4. Generate instructions
npx agentrc instructions --repo /path/to/repo --areas

# 5. Create eval scaffold
npx agentrc eval --init --repo /path/to/repo

# 6. Run evaluation (after editing agentrc.eval.json)
npx agentrc eval agentrc.eval.json --repo /path/to/repo

# 7. Create PR with all configs
npx agentrc pr owner/repo

# 8. For multiple repos
npx agentrc batch owner/repo1 owner/repo2 --output batch-results.json

Build & Test Commands

npm run build          # Compile TypeScript → dist/
npm run typecheck      # Type check
npm run lint           # ESLint
npm run test           # Single run
npm run test:watch     # Watch mode

Run without building:

npx tsx src/index.ts <command> [options]

Global Flags (All Commands)

  • --json — Machine-readable JSON to stdout
  • --quiet — Suppress progress output
  • --accessible — Screen reader friendly