AI_Global_Tools

This vault is the source of truth for reusable agent tooling across Claude Code and Codex.

Canonical resolution rule

Treat this absolute path as the master location regardless of the current repo or working directory:

• /mnt/hyperliquid-data/projects/obisidan/AI_Global_Tools

When an agent is asked to use a skill, workflow, prompt, agent definition, template, or MCP definition, resolve it from this vault first.

Lookup priority:

1. tool-specific canonical folder in this vault (claude/ or codex/)
2. shared canonical folder in this vault (shared/)
3. legacy repo-local copies only if the user explicitly asks for them or no canonical vault copy exists

Repo-local duplicates should be treated as mirrors, compatibility shims, or legacy copies unless explicitly declared authoritative by the user.

Canonical contents

• shared/prompts/system/
• shared/prompts/library/
• shared/agents/
• shared/skills/
• shared/templates/
• claude/skills/
• claude/prompts/system/
• codex/skills/
• codex/prompts/system/
• mcp/

Notable reusable pattern docs:

• LLM-Wiki-Pattern.md
• shared/templates/llm-wiki/
• shared/skills/herdr/

Safety boundary

This vault is intended for reusable tooling content, not secrets.

Keep these outside the vault:

• API keys
• auth tokens
• credential JSON files
• transient logs, history, and session state
• secret-bearing MCP environment values

Runtime bridges

The user home directory still contains the live runtime locations expected by the CLIs, but those locations now point back here where possible via symlinks and launch wrappers.