basalt-ai/Cobalt Eval

Run Cobalt AI agent experiments and post rich PR comments with results, comparisons, and AI analysis

View on GitHub

Trust Signals

Scorecard Score
not yet scored
Maintenance Recency
Activelast commit Feb 19, 2026
License
MIT

Pinned Snippet

workflow.ymlSHA-pinned
uses: basalt-ai/cobalt@802de59b8615f4622c6abd2b0626f8cf948a3692 # v0.3.0

tags can be moved; commit SHAs can't. why a SHA?

namedescriptionrequireddefault
experiment_filesExperiment file(s) to run (glob pattern or comma-separated). Uses cobalt.config defaults if empty.no""
working_directoryWorking directory for running experimentsno.
filterFilter experiments by name (case-insensitive substring match on filenames). Maps to CLI --filter flag.no""
concurrencyOverride concurrency setting (number of parallel experiment items). Maps to CLI --concurrency flag.no""
ciEnable CI mode with threshold validationnofalse
api_keyAPI key for LLM judges and AI summary. Supports both OpenAI and Anthropic (auto-detected from key prefix).no""
ai_summaryGenerate an AI-powered analysis of results (requires api_key)nofalse
github_tokenGitHub token for posting PR commentsno${{ github.token }}
comment_on_prWhether to post a comment on the PRnotrue
package_managerPackage manager to use: npm, pnpm, yarn, or auto (detect from lockfile)noauto
install_depsWhether to install dependencies before running experimentsnotrue
step_keyUnique key for comment deduplication. Change only if running multiple Cobalt steps in one workflow.no""
namedescription
passedWhether all experiments passed CI thresholds (true/false)
total_experimentsNumber of experiments that ran
summary_jsonJSON string of all experiment summaries