Upgrade agents or skills through a gated 5-phase pipeline with scoring.
npx skills add https://github.com/notque/claude-code-toolkit --skill agent-upgrade
Please help me install this Agent Skill. Skill: agent-upgrade Source: https://github.com/notque/claude-code-toolkit/tree/main/skills/agent-upgrade Command: npx skills add https://github.com/notque/claude-code-toolkit --skill agent-upgrade
This skill establishes a gated, objective upgrade workflow to improve an existing agent or skill.
Provide the target agent and run the agent-upgrade skill to start the AUDIT phase.
High-intent search queries and answers about installing and using this skill.
To systematically improve an existing agent's quality and template alignment, use a gated upgrade workflow that enforces a structured pipeline including auditing, diffing, planning, and re-evaluating with scoring. This ensures objective, reproducible enhancements.
A gated agent upgrade pipeline is a structured workflow that enforces user approval gates between phases like auditing, planning, and implementation. It works by scoring the agent before and after changes to guarantee quality improvements and template alignment.
Reproducible quality improvements when updating agent templates are ensured by applying a controlled, multi-phase pipeline with objective scoring and retro-education. This enforces structural, contextual, and safety requirements during every modification.
Yes, you can enforce plan gating and operator context during agent modifications by utilizing an objective upgrade workflow. This process requires explicit user approval before implementation, ensuring controlled changes and strict adherence to safety requirements.
The best way to evaluate an agent before and after making changes is to use a structured pipeline that applies scoring at the initial audit phase and a final re-evaluation phase, guaranteeing measurable quality improvements and template alignment.
az9713Initialize a project with dark code prevention directories, CLAUDE.md section, and baseline audit.
marinvchScore agent outputs against a multi-dimensional rubric with per-dimension and overall scores.
FFCfelps1Plan, execute, and ship software features with code review and CI/CD integration.
Dragoon0xCompile git metrics into structured weekly retro markdown files.
