# Inugamidev Ultrathink Ultrathink Core > UltraThink Core - Skill: `tomevault-io/inugamidev-ultrathink-ultrathink-core` (Agent Skill, multi-file: 2 files) - Install (CLI): `npx skillmds@latest add tomevault-io/inugamidev-ultrathink-ultrathink-core` - Raw SKILL.md: https://api.skillmd.com/api/skills/tomevault-io/inugamidev-ultrathink-ultrathink-core/raw - Safety review: pending (external: skill-scanner PASS, skillspector PASS) - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Coding & Dev Tools - Author: tomevault-io (https://skillmd.com/u/tomevault-io) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/tomevault-io/inugamidev-ultrathink-ultrathink-core --- # UltraThink Core Use this skill for non-trivial work in this repository or any project where UltraThink is installed. ## Contract 1. Read `AGENTS.md` for the active Codex instructions. 2. Check `.ckignore` before broad file exploration. 3. Use VFS/code-intel before full file reads when those tools are available. 4. Route to the full skill mesh only on demand: - registry: `.claude/skills/_registry.json` - skill bodies: `.claude/skills//SKILL.md` 5. Keep user/developer/system instructions above UltraThink when there is a conflict. 6. For substantial multi-step work, create or update a Markdown task ledger and mark tasks complete as they land. See `docs/protocols/task-ledger.md`. 7. Preserve unrelated worktree changes. 8. Before ending substantial work, write a `.handoff/handoff--.md` file. ## Why This Skill Is Small Codex has a startup skill-description budget. Do not expose every UltraThink skill directly in `.agents/skills`; that causes budget warnings and omitted skills. This facade keeps Codex startup small while preserving on-demand access to the complete `.claude/skills` mesh. --- > Source: [InugamiDev/ultrathink](https://github.com/InugamiDev/ultrathink) — distributed by [TomeVault](https://tomevault.io).