agent-instructions-setup

AI coding agent instruction file init and retrofit. Builds a single routing chain so every AI tool reads the same source: tool symlink -> AGENTS.md -> .agents/README.md (overview plus top-level directory routing table) -> each top-level <dir>/README.md -> detail files. Sets up AGENTS.md, idempotent symlinks across major AI tools (Claude Code, Cursor, Copilot, Windsurf, Cline, Roo Code, Gemini CLI, Codex, Zed, Antigravity, Amp, Aider, Continue), and a .agents/ meta directory (README, schema, conventions, preset.json, hooks/pre-commit-secrets.sh, local). No agents/ content directory and no team or hub mode branching - the chain is single-mode and AI-first. Bundles a meta-validator companion skill that auto-fixes frontmatter / relations / git timestamps, rebuilds .agents/.tag-index, scans for secrets, and enforces tight length recommendations (AGENTS.md 50/80/120, README.md 80/120/200). Use whenever the user wants to initialize a project, bootstrap AGENTS.md, unify instruction files across AI tools, retrofit an existing project, validate routing READMEs, or make a repo AI-navigable - even if they say 'AI rules', 'agent config', or 'check AGENTS.md'.