RAGit Agent Skill

use-ragit is the repository-managed canonical skill source for operating RAGit inside Codex, while keeping the main rules reusable for Claude and Gemini through shared reference files.

What It Solves

• Resolve the correct runtime before the first RAGit command
• Check repository readiness with status --format json
• Route agent intent to the right command
• Keep query, context pack, memory recall, memory wrap, memory promote, and hooks install from being mixed up
• Enforce --dry-run before mutating commands

Command Selection

• Use status for the first readiness check.
• Use doctor only after a failure or when you need structured diagnostics.
• Use describe before first-time command integration.
• Use query for raw retrieval hits.
• Use context pack for a prompt-sized, goal-bounded handoff packet.
• Use memory recall to resume active work from working memory.
• Use memory wrap to preserve session continuity.
• Use memory promote only for stable, searchable knowledge.
• Use hooks install only when automatic post-commit or post-merge ingest is explicitly requested.

Codex Install Path

Canonical source in this repository:

skills/use-ragit

Recommended Codex install target:

${CODEX_HOME:-$HOME/.codex}/skills/use-ragit

Typical symlink flow:

mkdir -p "${CODEX_HOME:-$HOME/.codex}/skills"
ln -s "/absolute/path/to/ragit/skills/use-ragit" \
  "${CODEX_HOME:-$HOME/.codex}/skills/use-ragit"

If you prefer not to symlink, copy the folder instead.

Shared References For Claude And Gemini

Claude and Gemini do not need a separate v1 wrapper pack. Reuse the same rules by pointing project instructions to these files:

• skills/use-ragit/references/command-routing.md
• skills/use-ragit/references/memory-discipline.md
• skills/use-ragit/references/payload-patterns.md
• skills/use-ragit/references/interop.md

That keeps command routing, safety rules, and memory discipline aligned across agents while leaving only metadata and trigger wording product-specific.