BlindSpot
tell your AI exactly which lines it forgot to test
npx -y github:Isco0819/blindspotEver run into this?
Codecov (acquired by Sentry, millions of CI integrations) proves devs obsess over test coverage — but there's no MCP interface for it. AI coding assistants write code without knowing which lines they just added have zero coverage, and the gap only surfaces in CI minutes later.
The fix
BlindSpot runs your existing test suite with coverage instrumentation and exposes the gap as MCP tools — your agent calls get_uncovered_lines(file) mid-session and self-corrects before finishing the task.
What it does
- get_uncovered_lines(file_path) returns exact line ranges with zero coverage so the agent writes targeted tests immediately
- get_coverage_diff() compares coverage before/after current git changes and surfaces only newly uncovered lines
- npx blindspot show — instant human-readable per-file coverage summary in terminal, zero config or account needed