Experiment · Workflows & capabilities / Code & architecture
Project Blueprint Agent Team
Follow six design roles from a project request to per-file implementation prompts.
Repository: CinvanaAI/project-blueprint-agent-team
Recorded synthetic example. An executable experiment using supplied synthetic inputs.
See the idea in action.
Six design artifacts and per-file implementation prompts.
A project request passed through six named roles.
{
"request": "Build a greeting tool",
"calls": [
{
"role": "prompt_interpreter",
"instruction": "Clarify scope, features, constraints, and assumptions. Return a structured project summary.",
"context_keys": [
"original_prompt"
]
},
{
"role": "system_architect",
"instruction": "Design a complete folder and file blueprint from the clarified summary.",
"context_keys": [
"original_prompt",
"project_summary.md"
]
},
{
"role": "dependency_mapper",
"instruction": "Explain dependencies and interactions among the proposed modules.",
"context_keys": [
"original_prompt",
"project_summary.md",
"folder_map.md"
]
},
{
"role": "file_prompt_designer",
"instruction": "Return strict JSON with a files array. Each item needs a relative path and implementation prompt.",
"context_keys": [
"original_prompt",
"project_summary.md",
"folder_map.md",
"dependency_map.md"
]
},
{
"role": "validator",
"instruction": "Check names, references, dependencies, completeness, and alignment with the original request.",
"context_keys": [
"original_prompt",
"project_summary.md",
"folder_map.md",
"dependency_map.md",
"file_prompts.json"
]
},
{
"role": "spec_writer",
"instruction": "Compile the validated result into a concise downstream handoff without claiming code was built.",
"context_keys": [
"original_prompt",
"project_summary.md",
"folder_map.md",
"dependency_map.md",
"file_prompts.json",
"validation_report.md"
]
}
],
"role_artifacts": {
"project_summary.md": "# Clarified project\n\nBuild: Build a greeting tool\n\nConstraint: keep the example offline and testable.",
"folder_map.md": "# Folder map\n\n- `app/core.py`: domain behavior\n- `tests/test_core.py`: offline verification",
"dependency_map.md": "# Dependencies\n\n`tests/test_core.py` imports `app/core.py`; the domain has no external service dependency.",
"file_prompts.json": "{\n \"files\": [\n {\n \"path\": \"app/core.py\",\n \"prompt\": \"Implement the clarified domain behavior.\"\n },\n {\n \"path\": \"tests/test_core.py\",\n \"prompt\": \"Test the public domain behavior offline.\"\n }\n ]\n}",
"validation_report.md": "# Validation\n\nThe two files agree on names, dependency direction, and offline scope.",
"final_summary.md": "# Handoff\n\nThe project blueprint is internally consistent and ready for implementation review."
},
"implementation_prompts": {
"project/app/core.py": "Implement the clarified domain behavior.\n",
"project/tests/test_core.py": "Test the public domain behavior offline.\n"
},
"software_implemented": false
}Try the example.
From the repository root, follow the dependency requirements in the README. This example uses supplied synthetic material.
python -m pip install -e . python -m blueprint_team "Build a greeting tool" --output ./demo-output
Six design artifacts and per-file implementation prompts.
Complete setup and instructions ↗The interesting part.
Separate role outputs make a proposed design traceable before code creation begins.
Where it came from.
An AutoGen role-based project design and implementation-prompt experiment.
Decouples the provider, removes fixed private inputs and bounds materialization; supplies offline stages.
Outputs include implementation prompts, not a completed generated application; no specific platform parent is established.