AI Agent Workflow Generator
Generate maintainable coding-agent instructions, skills and workflows.
Project setup
Build a maintainable coding-agent setup from project context. Review the workflow, checks and files before exporting.
Technology stack
Optional technology details
Adds an independent reviewer for larger repositories or monorepos.
Use commands your project actually declares. Generate setup refreshes the recommended checks; no command is executed here.
Workflow editor
Reorder steps, edit goals, connect branches, choose roles and attach skills. Every successful path must pass verification.
9 steps
- 1Explore repositoryAction
Next: Understand requirements
- 2Understand requirementsAction
Next: Decision
- 3DecisionDecision
Next: Plan / Implement
- 4PlanAction
Next: Implement
- 5ImplementAction
Next: Test
- 6TestAction
Next: Validate
- 7ValidateCheck
Next: ReviewOn failure: Implement
- 8ReviewReview
Next: Handoff
- 9HandoffHandoff
Next: Complete
03Generated files
Changed and conflicting files require explicit inclusion. Existing files are never overwritten by this tool.
Enter a project name.
New · AGENTS.md
Generated repository files use English for portability. Editing a file clears its acceptance; the manifest remains the canonical model.
Persistent context estimate
Approximation: one token per four characters. Counts global instructions only; referenced content adds context when loaded.
Reset draft
Related Tools
What is AI Agent Workflow Generator?
Generate a maintainable repository operating model for coding agents. A deterministic compiler turns project context into separate instructions, triggered skills, scoped rules, specialist agents, workflows and a portable agent-workflow.yaml manifest.
How to use
Start a new project or select a local folder or ZIP. Enter project details and real commands, choose coding-agent targets and workflows, edit steps and branches, then review generated files, explanations and diffs. Accept changed files explicitly before exporting. Save a named draft or import the manifest to continue editing.
Local analysis and verification
The bounded scanner detects stacks, package managers, declared commands with working directories, source structure and existing agent infrastructure. Findings cite evidence. Static checks flag missing verification, invalid connections, command mismatches, duplicate rules and context bloat. Every workflow has completion and failure recovery behavior.
Privacy and limitations
Repository contents are analyzed locally without uploads, command execution or an AI API. Sensitive paths and likely secret-bearing content are excluded. Drafts store the model but not repository bodies. Secret detection is heuristic, context tokens are approximate, and generated English configuration guides agents without enforcing runtime behavior. Reattach repositories after reopening drafts to verify current differences.
