chenkeyv/agent-toolbox
Reusable skills for agent-asset audits and upgrades, project setup, and reliable agent behavior.
Audit and modernize a repository's Markdown documentation and agent-facing instruction assets for concise, reliable, context-efficient use. Use when the user asks to review AGENTS.md, CLAUDE.md, skills, rules, prompts, memory guidance, MCP documentation, or a full documentation corpus for verbosity, duplication, conflicts, poor routing, outdated model assumptions, capability upgrades, or reconstruction opportunities. Produce evidence-backed advice without rewriting files unless the user asks for implementation.
Investigate a missed instruction, incorrect agent behavior, repeated mistake, or disappointing workflow outcome and turn it into a durable prevention. Use when the user asks why an agent did or did not do something, points out behavior that should not recur, says an instruction, skill, or rule was ignored, or asks how to prevent the same mistake. Compare expected and actual behavior using evidence, locate the failure layer, recommend the correct enforcement surface, and verify the fix without inventing hidden reasoning.
Safely rename a Git repository branch from master to main. Use for either a local-only branch rename or a full remote default-branch migration with upstream tracking, GitHub settings, branch protections, and optional remote master cleanup. Distinguish the requested scope before any remote mutation.
Set up, refresh, or repair a repository for agent-assisted work, especially skills-first Agent Toolbox workflows. Use when the user asks to configure or update a project for coding agents, add or refresh project instructions, install or reference Agent Toolbox, create project-owned memory or checkpoints, add project-owned skills, verify an existing setup, or make agent configuration idempotent.