Text & enterprise / FIUND
Terminal & software-agent task data.
Explore terminal task datasets for software-agent training and evaluation. Define reproducible environments, task instructions, reference solutions and verifiable success criteria before requesting availability.
Training & evaluation · Scope confirmed for your project
SOURCE CONTEXT
Source structure and relationships.
A task label or benchmark tier does not demonstrate quality. Inspect the executable environment, test the success criteria and review reference solutions. Separate training material from held-out evaluation tasks, document provenance and exclude credentials or private source code from any delivery.
Command-line task instructions
Sandboxed software environments
Reference trajectories and task checks
Software-agent evaluation · Tool-use training · Task completion assessment
TECHNICAL BRIEF
Define the source and structure.
These are decisions to agree for your collection, rather than specifications assumed across every source.
Environment versions and dependencies
Set the environment versions and dependencies requirements that matter to your task. Include an acceptable example and any exclusions to check during sample review.
Success criteria and reproducibility
Describe the beginning, intermediate steps and outcome a useful example should contain. Include interruptions or failed attempts if they matter to your task.
Code licensing and benchmark contamination
Set the code licensing and benchmark contamination requirements that matter to your task. Include an acceptable example and any exclusions to check during sample review.
BEFORE DELIVERY
A defined scope.
A considered handoff.
Evaluate the fit.
Agree what a useful sample must demonstrate, including the source conditions and required relationships.
Confirm the permissions.
Review the intended AI uses, relevant exclusions and documentation for the selected material.
Agree the package.
Specify files, metadata, preparation and acceptance criteria before proceeding with delivery.
START A CONVERSATION
Let’s scope your
data request.
Tell us what your model needs from terminal & software-agent task data. We’ll assess the sourcing options and discuss a suitable next step.
Availability and collection terms are confirmed after review.
Frequently asked questions
Can relationships between documents or assets be preserved?
Include the identifiers, links or source-to-output pairings your model needs. Privacy treatment, file structure and permitted uses are assessed for the proposed material.
What should I include in a terminal & software-agent task data brief?
Start with your model task and the source material you need. Include environment versions and dependencies, success criteria and reproducibility, code licensing and benchmark contamination so we can assess fit and propose a useful evaluation sample.
When are availability, pricing and permissions confirmed?
After we assess your brief and identify suitable material or a capture scope. Samples, permitted AI uses, preparation and delivery terms are agreed for the proposed collection before a purchase.