The real benchmark for a dead-code scanner is not "does it pass my toy suite?"
Its -> will maintainers of Black, NetworkX, mitmproxy, pypdf, Optuna, etc merge the cleanup PR?
For Skylos so far:
9 merged PRs
46 files changed
372 lines deleted
Skylos isnt an alternative to GitHub Actions. It runs in gh actions. The real question is .. "what runs in your pipeline." GitHub's native scanning covers SAST + deps + secrets (and CodeQL costs money on private repos). Skylos adds the things GitHub doesn't really do (or u can configure it if you need but there's setting up tax etc).. Things such as dead code, AI-code hallucinations, cicd config security, LLM guardrails in one local-first CLI you can also run before you ever push. Also we have published benchmarks showing that we have beaten other open source libraries in terms of accuracy and recall
@owenthcarey Working on a tool to catch vibe coded vulnerabilities and quality regressions using both agents and static findings
https://t.co/5BJ6A5M8dY
I’m thinking of running a small Skylos OSS challenge.
Prize pool: USD $300 3 winners × $100 each
The idea is simple: show me what you wish Skylos had that will make you use it.
It can be:
- a new detector/rule
- better framework support
- a workflow improvement
- uiux and docs improvements
- improving trust
I'm not expecting full production features. I mostly want to see what contributors would actually want to use, and what would make Skylos more useful in real projects.
Submit a PR during the challenge window. I’ll pick the top 3 accepted contributions based on usefulness, clarity, code quality, and fit with Skylos.
Paid via PayPal after winners are announced. Would anyone be interested?