greynewell/SWE-bench Pro Evaluation

Run SWE-bench Pro preflight validation and agent evaluation using mcpbr

View on GitHub

Trust Signals

Scorecard Score
not yet scored
Maintenance Recency
Stale
License
None
namedescriptionrequireddefault
modeOperation mode: 'preflight' (validate golden patches) or 'evaluate' (run agent evaluation)nopreflight
benchmarkBenchmark name (e.g., swe-bench-pro, swe-bench-lite)noswe-bench-pro
sample-sizeNumber of instances to evaluate (empty = all)no""
task-idsComma-separated task instance IDs to runno""
filter-categoryFilter by language or repository substringno""
max-concurrentMaximum concurrent Docker containersno2
timeoutPer-test timeout in secondsno300
fail-fastStop on first failurenofalse
configPath to mcpbr YAML config file (evaluate mode)no""
anthropic-api-keyAnthropic API key (evaluate mode)no""
modelModel override (e.g., claude-sonnet-4-5-20250514)no""
output-formatComma-separated output formats: json, junit, markdown, htmlnojson,junit
mcpbr-versionmcpbr version to install (e.g., 0.14.1, or empty for latest)no""
shard-indexShard index for parallel runs (0-based)no""
shard-totalTotal number of shards for parallel runsno""
namedescription
results-pathPath to the results directory
totalTotal number of instances evaluated
passedNumber of instances that passed
failedNumber of instances that failed
success-rateSuccess rate as a percentage