← Back to the Hub
Preset
benchmark-coding
benchmark-coding
Autonomous coding harness for headless benchmark runs: never asks the user, investigates exhaustively, always attempts and verifies a fix.
builtinOpen sourceReadyContext replay sessions
Specification
| Category | Model and task presets |
|---|---|
| Runtime | builtin |
| Interface | TUI and CLI |
| Source | built-in-template |
| Licence | Apache-2.0 |
| Continuity | Context replay |
| Support | Ready |
| Based on | benchmark-coding |
Runtime behaviour
Capabilities
single workflowStructured run events
Tools
read_filewrite_filelist_directoryedit_fileinsert_textpatchmulti_editgrepgloblocal_code_searchrepo_searchcode_searchbashtodo_writetodo_readPolicies and ownership
- +Sandbox: local
- +Approvals: yolo
- +Shell: allowed
- +Network: blocked
- +Checks: enabled
Get started
Inside SuperQode
$ :harness switch benchmark-coding$ :harness show benchmark-coding$ :harness statusFrom your shell
$ superqode harness run benchmark-coding "Describe your task"$ superqode harness show benchmark-codingEvaluate and optimize
Measure
$ superqode harness test --spec harness.yaml$ superqode harness eval --spec harness.yaml --tasks eval-tasks.yaml$ superqode harness bench --manifest harnessbench.yamlImprove
$ superqode harness optimize-omni --spec harness.yaml --tasks eval-tasks.yaml$ superqode harness optimize --spec harness.yaml --tasks eval-tasks.yaml$ superqode harness promote stage