74
/ 100
Good shape overall. A few tweaks would push it into the top tier.
Data recipes and robust infrastructure for training AI agents
Top fixes
Highest-impact changes first, ranked by point weight
- 1Tests18pt
Wire your tests to a documented command (e.g. a test script in your build config) so the suite is reproducible.
- 2CI/CD14pt
Add a step like `run: npm test`, `run: pytest`, or `run: tox` to your workflow file.
- 3CI/CD14pt
Add `pull_request:` to the workflow `on:` triggers.
- 4CI/CD14pt
Add a lint step to catch style issues automatically.
Working through the fixes? Let every push regrade itself.
The free GitHub App rescans this repo on every push and posts the grade as a commit check, so the score climbs without coming back to rescan by hand.
Scorecard
Every check, grouped by category and sorted worst-first
Documentation
97
Contributing guide is too short for full depth credit (−6 pts). 400+ words earns the full +12 pts.
→ Add setup instructions, code style notes, and how to run tests.
README is present.
README documents how to install the project.
Licensed under Apache-2.0.
Engineering
51
No linter or formatter configured.
→ Add [tool.ruff] to pyproject.toml. Ruff handles linting and formatting in one tool and replaces flake8, isort, and black.
No issue or PR templates found (−100 pts).
→ Add .github/ISSUE_TEMPLATE/ with bug_report.md and feature_request.md to guide contributors. It dramatically improves issue quality.
CI is configured (.github/workflows/submodules-nightly.yml).
Test files detected (data/all_puzzles/test_generate.py).
Lockfile present (uv.lock). Installs are reproducible.
Project health
96
Dependency manifest found (pyproject.toml).
Repository has a description.
Actively maintained (pushed within the last month).
.gitignore present.
Repository health signals
Activity, community, and responsiveness at scan time
Activity
- 379 / 694Commits (30d / 90d)
- 31Forks
- 0Releases
Community
- 50% - FairCommunity health
- 1 bus factorlowauthor own >50% of commits
- 233Watchers
Responsiveness
- 8d 24hMedian issue response
- 5d 5hMedian PR merge time
- 4Open issues
Repository files33 root entries
- .claude
- .githubGood: CI is configured (.github/workflows/submodules-nightly.yml).
- assets
- baselines
- chat_templates
- dataGood: Test files detected (data/all_puzzles/test_generate.py).Good: Environment pinned via data/frontiersmith/templates/Dockerfile.
- database
- docker
- docs
- eval
- hpc
- notebook
- notes
- prm
- rl
- schema
- scripts
- sft
- tests
- .agents
- .beta9ignore
- .dockerignore
- .gitignoreGood: .gitignore present.
- .gitmodules
- build_support.py
- CLAUDE.md
- CONTRIBUTING.mdIssue: Contributing guide is too short for full depth credit (−6 pts). 400+ words earns the full +12 pts.Fix: Add setup instructions, code style notes, and how to run tests.Good: Contributing guide includes setup/install instructions.Issue: Contributing guide lacks a code style section (−8 pts).Fix: Describe your linting/formatting rules and how to run them.Issue: Contributing guide lacks a testing section (−8 pts).Fix: Show contributors how to run the test suite (e.g. npm test, pytest, cargo test).Good: Contributing guide describes the PR/review workflow.Good: Contributing guide includes code examples.
- err.txt
- LICENSE.txtGood: Licensed under Apache-2.0.
- pyproject.tomlGood: Dependency manifest found (pyproject.toml).
- README.mdGood: README is present.Good: README is well structured with multiple sections.Good: README includes screenshots or visuals. Great for first impressions.Good: README has code examples.Good: README links to a live demo or deployed app.Good: README includes status badges.Good: README documents how to install the project.Good: README documents how to run the project.
- setup.py
- uv.lockGood: Lockfile present (uv.lock). Installs are reproducible.
Add this badge to your README
It updates automatically each time the repo is re-graded.
[](https://www.repo-grade.com/report/open-thoughts/openthoughts-agent)