Evaluate, shrink, or scaffold mentat skills with pass-rate and token-ratio gates.
npx skills add https://github.com/matheushenriquefs/mentat --skill mentat-skill
Please help me install this Agent Skill. Skill: mentat-skill Source: https://github.com/matheushenriquefs/mentat/tree/main/.agents/skills/mentat-skill Command: npx skills add https://github.com/matheushenriquefs/mentat --skill mentat-skill
The mentat-skill helps manage and streamline the process of working with mentat skills by providing functionalities like evaluation, shrinkage, and scaffolding.
Use the /mentat-skill eval command followed by the skill name to evaluate an existing skill.
High-intent search queries and answers about installing and using this skill.
To evaluate a mentat skill, run functional and cognitive evaluations against the skill using the eval command. This process scores the skill's performance while guarding results with pass-rate and token-ratio gates.
You can scaffold a new skill by generating a skill directory from pre-configured templates. This creates the necessary directory structure and configuration files required to start developing a new mentat skill.
Skill shrinkage analyzes existing SKILL.md files and proposes leaner versions. It reduces token usage while preserving core functionality, helping parallel coding agents process instructions more efficiently.
Yes, this tool handles parallel coding agent management with a focus on code rebase, re-gate, scoring, and landing operations. It coordinates multiple agents working simultaneously on the same codebase.
Pass-rate and token-ratio gates are threshold metrics used during functional and cognitive evaluations. They determine whether a skill meets performance standards by measuring test success rates and token consumption efficiency.
No external dependencies are required to use mentat-skill. It operates independently using its built-in scripts and references components to perform evaluations, shrinkage, and scaffolding tasks.
Kilo-OrgScaffold SKILL.md frontmatter and starter directories for new skills.
rjschieAnalyze branch commits and propose an optimized rebase order.
Marshall-HallenbeckCoordinates multiple AI agents in tmux split panes for parallel debugging and hypothesis exploration.
gelowzzzCreate, validate, and package Claude Skills into .skill files.
