DOS (dos-kernel)
Verifies AI agent claims by checking real git history and artifacts instead of trusting reports.
View on GitHub Updated 2026-06-15
What is DOS (dos-kernel)?
DOS acts as an independent referee for AI agents working on code repositories. Rather than accepting an agent's announcement that a task is finished, it examines concrete evidence such as commit history and file state to determine the true outcome.
The system supports both single-agent checks and multi-agent environments where several agents operate on the same codebase. It identifies conflicts, repeated work, or unsubstantiated claims by comparing reported progress against observable changes.
It targets developers and teams running agent fleets in CI pipelines or local setups who need reliable gates on agent output without adding complex configuration.
What you can build with DOS (dos-kernel)
Single agent verification
Run dos verify after an agent finishes a task to confirm the claimed features match actual commits before accepting the result.
Multi-agent fleet coordination
Monitor several agents working in parallel on one repo to detect overlapping changes or false reports that could corrupt shared code.
CI pipeline gating
Add DOS checks to automated workflows so merges or deployments only proceed when verification exits with a success code.
Install DOS (dos-kernel)
Install
pip install dos-kernelQuick start
> pip install dos-kernel
> dos init --hooks auto # finds the agent runtime(s) you already use, wires in the checks
>- 1Install the package with pip install dos-kernel.
- 2Navigate to your git repository root.
- 3Run dos init --hooks auto to connect the verification steps to your existing agent setup.
- 4Trigger an agent task and follow it with dos verify on the relevant claim identifiers.
- 5Use the returned exit code to decide whether to accept the agent's completion or reassign the work.
DOS (dos-kernel): pros & cons
Pros
- +Provides objective verification based on git and file evidence rather than agent statements.
- +Requires zero manual configuration on standard git repositories.
- +Works with existing agent runtimes through simple hook installation.
- +Scales from single checks to full fleet oversight with consistent exit-code results.
Cons
- –Primarily tied to git history so non-git projects need additional setup.
- –Limited to Python environments for installation and execution.
- –Verification scope depends on what artifacts the agent actually modifies.
Did you find this helpful?
Frequently asked questions
It inspects git commits and related artifacts to match the agent's claim against actual changes, returning SHIPPED or NOT_SHIPPED with corresponding exit codes.
User reviews
Verified reviews from the community shape this listing's rating.
Loading reviews…
Similar agents
Other general-purpose options worth comparing.
