npx claudepluginhub epicrunze/jarvisJaRVIS: Journaling As Recurrent Versioned Identity Sculpting. Post-task reflection, memory sculpting, and self-evolving agent identity — stored as markdown in your repo.
epicrunzeJournaling As Recurrent Versioned Identity Sculpting
AI agents forget everything between sessions. JaRVIS gives your agent persistent memory, post-task reflection, and a self-evolving identity — stored as plain markdown files you can read, edit, and version-control.
JaRVIS adds these skills to your AI coding agent:
/jarvis-init — One-time setup to scaffold the data directory (RUN THIS AFTER INSTALLATION)/jarvis-reload — Reload identity and memories mid-session (automatic on platforms with hooks)/jarvis-reflect — Post-task reflection that captures lessons and updates memories/jarvis-identity — Evolve the agent's identity based on accumulated experienceInstall JaRVIS as a Claude Code plugin:
/plugin marketplace add epicrunze/JaRVIS
/plugin install jarvis@jarvis-marketplace
Then run /jarvis-init to scaffold the JaRVIS data directory.
Paste this prompt into your AI coding agent to install or update JaRVIS:
Install or update JaRVIS (https://github.com/epicrunze/JaRVIS) in this project. Follow these steps exactly:
1. Download and extract the JaRVIS repo:
curl -sL https://github.com/epicrunze/JaRVIS/archive/refs/heads/main.tar.gz | tar xz
2. Detect the platform and set SKILLS_DIR:
- If .claude/ exists → SKILLS_DIR=".claude/skills"
- If .cursor/ exists → SKILLS_DIR=".cursor/skills"
- If .github/ exists → SKILLS_DIR=".github/skills"
- If .agent/ exists or AGENTS.md exists → SKILLS_DIR=".agent/skills"
- If none match, ask me which platform I'm using and where skills should be installed.
3. Create the skills directory if needed: mkdir -p "$SKILLS_DIR"
4. Copy the skill folders:
cp -r JaRVIS-main/skills/* "$SKILLS_DIR/"
5. Clean up: rm -rf JaRVIS-main
6. If JaRVIS hasn't been set up for this project yet, run /jarvis-init to complete setup. If JaRVIS is already set up, skip this step — the update is complete.
Works with Claude Code, Cursor, GitHub Copilot, Antigravity, and other AI coding agents. See install/PROMPT.md for details.
Other platforms: If your agent doesn't match any of the detected platforms, the install prompt will ask you where to put skills.
/jarvis-initwill then ask for your instruction file path.
Copy skills into your platform's skills directory (e.g., .agent/skills/ or wherever your platform loads skills from):
cp -r skills/* <your-skills-directory>/
Then run /jarvis-init to scaffold the JaRVIS data directory. The init skill will detect your platform and configure things automatically.
/jarvis-init — this scaffolds the JaRVIS data directory and configures your platform/jarvis-reflect — writes your first reflection/jarvis-reload manually on other platforms)/jarvis-reflect after completing tasks/jarvis-identity evolves the identity document/jarvis-validate to check artifact formatting and /jarvis-search to find past entriesflowchart TD
A["🔄 /jarvis-reload"] --> B["⚡ Work"]
B --> C["📝 /jarvis-reflect"]
C --> B
C -->|every 5 reflections| D["🧠 /jarvis-identity"]
D --> B
subgraph store ["📂 ~/.jarvis/projects/<slug>"]
J[("📓 Journals")]
M[("💾 Memories")]
I[("🧬 Identity")]
end
A -. reads .-> I
A -. reads .-> M
A -. reads .-> J
C -. writes .-> J
C -. updates .-> M
D -. reads .-> M
D -. reads .-> J
D -. evolves .-> I