dsh-tool-radar

The most painful moment with an open-source contribution agent is burning an entire session only to discover the target repo simply does not merge external PRs—maintainers go months without a single reply, the queue has hundreds of stale entries, and every merge is an internal one. dsh-tool-radar steps in before the agent starts working and delivers a clear, evidence-backed answer.

It exposes a single model-facing call, repo_radar, that returns a 0-to-100 score with itemised evidence: median days from open to merge on recent PRs, open-queue size and stale ratio, median time to the first maintainer comment, and the actual merge rate for outside contributors (NONE/CONTRIBUTOR/FIRST_TIMER). Every scoring rule carries a plain-language note plus machine-readable risk flags such as stale_pr_backlog or slow_maintainer_response. Scores of 70+ are labelled Contribute, 45-69 Consider, and below 45 Avoid. API usage is capped at five calls, unauthenticated by default; a token, when supplied, only raises the rate limit.

It is built for developers who run dsh-powered open-source workflows: whether you need an agent to shortlist upstream projects worth contributing to, or you want a quick health check before a batch of PRs, repo_radar turns a gut-feel decision into one grounded in live data.