← ClaudeAtlas

engineering-scorecardlisted

Score a company's developers and its most complex features out of 100 from git history, across any number of local repos and GitHub repos (read through the API, no clone), and produce a ranked report with charts. Use this whenever someone wants to know who built what, which developer did the hardest work, which features are most complex, how contribution splits across repos, or wants an engineering team scorecard, contributor analytics, developer leaderboard, codebase ownership map or "complexity by developer" report, even if they don't say "scorecard". Also use it to refresh or extend such a report with more repos.
Slasher190/engineering-scorecard · ★ 0 · Code & Development · score 72
Install: claude install-skill Slasher190/engineering-scorecard
# Engineering scorecard Turns git history into two rankings, each out of 100: - **Developers**, on eight weighted parameters: difficulty, hard-feature output, volume, breadth, ownership, tests, consistency and commit hygiene. - **Complex features**, on technical depth, domain risk and integration surface (which you rate by reading the code), plus measured size and tests, with the lead developer of each. The heavy lifting is in `scripts/scorecard.py` (standard-library Python). Your job is the judgment the script can't make: which repos belong together, which git names are one person, how hard each module is, and which features matter. Those choices decide whether the result is fair, so spend your effort there. `references/methodology.md` has every formula and the rubrics for tiers and feature ratings. `references/config.md` has the config schema. Read both before step 4. ## Be straight about what this measures The scores describe the git record: what was written, where, how hard that area is, and how it was committed. They don't see code quality, design, reviews, mentoring, incidents handled or business impact. Say so when you present results, and steer away from using the ranking alone for pay or performance decisions. The report carries the same caveats. ## Requirements `python3` (3.8+) and `git`. For GitHub repos that shouldn't be cloned, the `gh` CLI signed in with read access. Nothing is written outside the output folder, and nothing is sent anywhere except read-o