Guard Skills catches coding agent failure modes
guard-skills is an open-source quality-assurance suite designed to catch systematic failure modes in AI-generated code, tests, and documentation. By acting as a secondary local review pass for AI coding agents like Claude Code or Cursor, it targets common agent mistakes such as hollow tests, silent catch blocks, hallucinated APIs, and stale comments before they are committed or merged into production.
As AI coding agents become mainstream, their propensity to write silent catch blocks or hollow tests makes automated post-processing tools like guard-skills necessary to enforce code quality.
* Proactive validation: It addresses AI-specific vulnerabilities, such as mock abuse in test suites and API hallucinations.
* Local and secure: Running locally means it does not incur additional API costs or leak source code.
* Workflow integration: Serves as a vital quality gate between agent generation and final human review.
DISCOVERED
2h ago
2026-06-17
PUBLISHED
2h ago
2026-06-17
RELEVANCE
AUTHOR
Github Awesome