Auditar Skills: validate Claude Code and Codex skills
Checks skills against 46 rules, highlights the five weakest and lets you test them with and without a validation hook.
Auditar Skills is an open toolkit for checking Claude Code and Codex skills. Skills are folders with a SKILL.md file that guides an agent to perform a task the same way each time. The toolkit combines skill-creator-plus, from RoboNuggets, with an INEMA guide to finding problems and reviewing instructions.
The validator checks each skill against 46 rules and points to problems by file and line. It only reads files and requires no pip installation. The guide helps prioritize the five skills with the most problems, generate a report, and test changes with and without the skill. Changes are made only with the user's approval.
To get started, you need Python 3.8 or later and a copy of the repository. The validator can run without an agent; the skill's audit, creation, and refinement modes require Claude Code or Codex. The validation hook is optional and must be configured: it runs when a SKILL.md is saved and reports errors for correction.
The link opens the original content; some require a subscription.
Go further on INEMA.CLUB
More news
- 05/10Tríade Letal: could your AI agent leak your data?A checklist and MCP inventory to help identify data leak risks from AI agents.
- 05/10SIL Loop R 1.1: adopted rules carry over to the next sessionPromotes rules to a monitored block in AGENTS.md and flags when it is outdated or a rule leaks.
- 05/10Which AI draws a capybara on a bicycle best?A guide presents a competition between AIs to draw a capybara on a bicycle.
- 05/10filmarca: turn your website into your brand filmCreates a script and scenes with subscription AI for review before rendering, in 16:9, 9:16 and 1:1, in PT, EN and ES.
- 05/10Atende Nutrição: meal plans and alerts on WhatsAppA plan kit for tracking meals with a food diary and health alerts on WhatsApp; includes 99 tests and /goal.