:rocket: <https://github.com/seqra/opentaint|OpenT...
# feed
a
🚀 OpenTaint 0.4.0 is out! The release adds a set of AI agent skills for application security. Most static analyzers builtin rules are generic, so they miss the bugs that are specific to your project. These skills close that gap. The agent learns your project and does the expensive work once: • maps the attack surface and writes project-specific rules • models your library data flows that generic rules don't cover • triages findings and confirms the real ones dynamically at runtime Then the deterministic engine reuses those artifacts to catch the project-specific dataflow bugs (SQL injection, SSRF, RCE) on every commit, CPU only, no model calls. On a 130K-line Kotlin project: the first deep run writes project-specific rules for 3 ours, then every commit after scanned in under 2 min on CPU alone. The OpenTaint is an open source, with first-class Kotlin support. Runs in Claude Code, OpenCode, Codex, and other agent tools:
npx skills add <https://github.com/seqra/opentaint>
Write-up: https://opentaint.org/blog/appsec-agent/
K 3
👍 1