Dorcha releases free GitHub Action 'deslop-kitten' that flags AI slop pull requests before review

A free GitHub Action that scans pull request diffs for malicious code like command execution, credential access, exfiltration, auto-run hooks, obfuscation, and prompt injection. It uses no LLMs — instead a deterministic regex + semgrep layer meant to run before AI code reviewers like Greptile or CodeRabbit, which it can then trigger. It addresses the problem of well-hidden malicious code slipping past quality-focused AI review, targeting OSS maintainers.

Products·launched

Driftwatch — Cursor rules + GitHub Action that fails PRs when agents skip or weaken tests

Driftwatch pairs Cursor rules with a GitHub Action that fails PRs when AI agents skip tests, weaken assertions, or make out-of-scope changes. LLM coding agents like Cursor are the thing being policed; the Action itself is Python stdlib. It targets teams whose Cursor-generated PRs quietly erode CI discipline.

Driftwatch pairs Cursor rules with a GitHub Action that fails PRs when AI agents skip tests, weaken assertions, or make out-of-scope changes. LLM coding agents like Cursor are the thing being policed; the Action itself is Python stdlib. It targets teams whose Cursor-generated PRs quietly erode CI discipline.

Products·building

Scanity — codebase security vulnerability scanner for AI-generated code

Scanity scans codebases to find security vulnerabilities, aimed at bugs that slip through when AI writes code quickly. It integrates with GitHub Actions to check PRs before merge with suggested fixes, and can run full repo scans. LLMs are used to analyze code and generate fixes; it builds containerized test units to validate findings.

Scanity scans codebases to find security vulnerabilities, aimed at bugs that slip through when AI writes code quickly. It integrates with GitHub Actions to check PRs before merge with suggested fixes, and can run full repo scans. LLMs are used to analyze code and generate fixes; it builds containerized test units to validate findings.

PR-Agent, an open source AI tool for code review, now offers context retention across pull requests, enabling it to track discussions and changes over time. This improves review consistency and reduces repetitive feedback for developers.