Library · Workflows & capabilities
Materialized Task Runner
Run a saved, trusted Python task in a fresh interpreter and keep the source and result evidence.
Repository: CinvanaAI/materialized-task-runner
Recorded synthetic example. Source with a complete offline example.
See the idea in action.
A fresh interpreter returns {"greeting":"Hello Avery"}; source and result hashes remain in evidence.
A formatter package, run_workflow function and {"name":"avery"}.{
"kind": "captured synthetic execution projection",
"command": "python -m examples.demo",
"input_example": "examples/demo.py",
"observed_output": {
"unauthorized_refused": true,
"status": "complete",
"result": {
"greeting": "Hello Avery"
},
"runtime_cleaned": true
},
"projection_note": "Selected fields from a successful local run. Machine paths, temporary identifiers and execution timestamps are omitted. Re-run the linked example for fresh evidence."
}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 examples.demo
A fresh interpreter returns {"greeting":"Hello Avery"}; source and result hashes remain in evidence.
Complete setup and instructions ↗The interesting part.
Separating task materialization from the authoring workbench makes saved work independently executable and reviewable.
Where it came from.
Task materialization and execution subsystem of historical Python Agent Foundry.
Moves from shared running directories to unique runs, explicit authorization, bounds and durable evidence.
The focused runner is adapted for independent use; it does not replace the parent source.
Follow the family: Builder, Factory and the Python Foundry
- Extracted fromPython Agent Foundry Workbench
See the integrated setting from which this independently useful mechanism was separated. The standalone boundary has documented public adaptations; this is not a promise of current drop-in interchangeability.
Source ↗