gald3r
Health Uyari
- License — License: NOASSERTION
- Description — Repository has a description
- Active repo — Last push 0 days ago
- Low visibility — Only 8 GitHub stars
Code Uyari
- network request — Outbound network request in platforms/amp/.agents/skills/g-skl-crawl/scripts/crawl_url.py
Permissions Gecti
- Permissions — No dangerous permissions requested
This framework provides persistent, file-based memory, multi-repo orchestration, and adversarial quality gates for AI coding agents. It acts as an infrastructure layer that helps various AI assistants remember context and coordinate tasks across different workspaces and sessions.
Security Assessment
The automated code scan reviewed 12 files and found no dangerous patterns, hardcoded secrets, or requests for risky permissions. The tool relies on file-based memory to persist data across sessions, which means it will read from and write to your local file system. Because it orchestrates coding agents, it inherently facilitates code execution and system changes via the connected agents (like Claude Code). Overall risk: Low.
Quality Assessment
The project is licensed under the permissive and standard MIT license. It is actively maintained, with the most recent code push occurring today. However, community trust and adoption are currently quite low, with only 17 GitHub stars. This suggests it is an early-stage or niche project.
Verdict
Safe to use, though you should expect an early-stage project experience given its small community footprint.
AI dev framework: 22 agents, 100 core skills + 142 addon skill packs, 149 commands. Runs on 21 AI tools: Cursor, Claude Code, Copilot, Windsurf, Cline, Aider, Roo, Kiro, OpenHands, Gemini, Codex, Goose, Warp & more. File-based memory persists across sessions.
gald3r — AI Agent Framework for Your Project
File-based memory, task management, and agent orchestration that installs in minutes —
now backed by a bundled file-first engine (a gald3r CLI + MCP server,
zero LLM calls). Works in Cursor and Claude Code (Tier 1), plus
34 AI coding platforms — no server, no database, no Docker.
Changelog |
Contributing |
All 34 platforms
What is gald3r?
gald3r is a template you drop into any project to give your AI coding assistant a persistent brain.
Once installed, your AI gains:
- Persistent memory across sessions — tasks, bugs, plans, constraints survive every restart
- 110 skills for common dev workflows (code review, QA, task management, planning, and more)
- 177 commands invoked directly in chat (
@g-status,@g-go,@g-task-new,@g-bug-report) - 37 hooks that fire on IDE events (session start, file save, commit)
- 12 rules that keep the agent disciplined every session
- Works in both Cursor and Claude Code over one shared
.gald3r/brain — plan in one, code in the other - A bundled file-first engine (Mode-A, new in 2.0) — every system (tasks, bugs, vault, releases, …) is driven by a deterministic Python core via the
gald3rCLI or an MCP server, with zero LLM calls.gald3r doctorkeeps the install healthy. One prerequisite:uv.
Everything is plain markdown files in your repo. No accounts, no API keys beyond what you already have.
Quick Install
Option 1 — Copy the template (recommended)
git clone https://github.com/wrm3/gald3r.git
# Default: installs Cursor + Claude Code + shared brain
cp -r gald3r/project_template/. /path/to/your/project/
Then open your project in Cursor or Claude Code and run @g-setup / /g-setup.
Option 2 — Installer script (supports all 34 platforms)
# Default: Cursor + Claude Code (same as copying project_template/)
.\setup_gald3r_project.ps1 -TargetPath "C:\MyProject"
# Install for a specific platform only
.\setup_gald3r_project.ps1 -TargetPath "C:\MyProject" -Platform windsurf
.\setup_gald3r_project.ps1 -TargetPath "C:\MyProject" -Platform cline
.\setup_gald3r_project.ps1 -TargetPath "C:\MyProject" -Platform cursor # Cursor only (no .claude/)
What Gets Installed
Default install (Cursor + Claude Code):
your-project/
├── .cursor/ ← Cursor config (rules, skills, commands, hooks, agents)
├── .claude/ ← Claude Code config (same skill set, markdown format)
├── .gald3r/ ← Shared project memory (tasks, bugs, plans, constraints)
├── .gald3r_sys/ ← gald3r system files (skills engine, platform specs)
├── AGENTS.md ← Universal agent instructions (read by both IDEs)
├── CLAUDE.md ← Claude Code entry point
└── WORKFLOW.md ← Project workflow definition
Platform-specific install (e.g. -Platform windsurf): same shared brain, plus the platform's config folder (.windsurf/rules/ etc.). Cursor and Claude config are skipped.
The engine (CLI + MCP)
New in 2.0: a bundled, file-first Python engine drives every system deterministically — no LLM,
no network, no Docker. It lives in .gald3r_sys/engine/. The only prerequisite isuv; the first run provisions it automatically.
Command line
# from your project root — the first run builds the engine (a few seconds), then it's instant
uv run --project .gald3r_sys/engine gald3r doctor # health check (add --fail-below 80 for CI)
For brevity, alias the prefix — alias gald3r='uv run --project .gald3r_sys/engine gald3r' — then:
gald3r task new --title "Wire up auth" --priority high
gald3r bug new --title "Login 500 on empty cart" --severity high
gald3r goal add --text "Ship the MVP by Friday"
gald3r vault ingest --title "JWT notes" --type article --source https://example.com/jwt
gald3r inbox # absorb staged task/bug drafts into live state
gald3r prompt get role.code_reviewer # load a reasoning brief
gald3r --json task list # machine-readable output
(python -m gald3r … works too, if you'd rather not alias.)
As an MCP server
Expose the same operations as ~20 MCP tools to any MCP-capable agent. Add to your client's MCP config:
{
"mcpServers": {
"gald3r": {
"command": "uv",
"args": ["run", "--project", ".gald3r_sys/engine", "gald3r", "mcp"]
}
}
}
Your agent then calls gald3r_task_new, gald3r_bug_list, gald3r_prompt_get, … directly.
No engine? Still works.
The engine is additive. The .gald3r/ state is plain markdown, and every slimmed skill keeps aSKILL.full.md fallback — so a files-only install (no uv / Python) runs unchanged.
Platform Support
| Platform | Tier | Notes |
|---|---|---|
| Cursor | ✅ Tier 1 | Rules (.mdc), skills, commands, hooks, agents |
| Claude Code | ✅ Tier 1 | Rules (.md), skills, commands, hooks, agents |
| Windsurf, Cline, Roo, Aider | ⚠️ Tier 2 | Rules + brain + AGENTS.md |
| Copilot, Codex, Augment, Gemini, Qwen, Continue | ⚠️ Tier 2 | Rules + brain + AGENTS.md |
| 20+ more | 🔜 Tier 3 | Brain + AGENTS.md (rules where supported) |
Use -Platform <name> with the installer, or copy platforms/<name>/ manually.
See gald3r_supported_platforms.html for the full list.
Cursor + Claude Code users get the full experience. Other platforms receive the shared
.gald3r/brain andAGENTS.mdinstructions, with rules where the platform supports them.
How It Works
Your project root
├── AGENTS.md ─────────────────────────────────────────────────────────┐
├── .cursor/ (rules + skills + commands) ← Cursor reads these │
├── .claude/ (rules + skills + commands) ← Claude Code reads these │
│ │
└── .gald3r/ ────────────────────────────────────────────────────────────┘
TASKS.md ← shared task list, visible to both IDEs
BUGS.md ← shared bug tracker
PLAN.md ← shared strategy & milestones
CONSTRAINTS.md ← rules the agent must never break
Every command you run in Cursor or Claude Code reads and writes these same files. Switch between tools anytime — context is never lost.
Key Commands
| Command | What it does |
|---|---|
@g-setup / /g-setup |
Initialize gald3r in a new project |
@g-status / /g-status |
Show project health: tasks, bugs, open items |
@g-go / /g-go |
Start an autonomous work session on the next task |
@g-task-new |
Create a new task with spec |
@g-bug-report |
File and triage a bug |
@g-medic |
Run self-diagnostics on the gald3r installation |
@g-plan |
Update and review the project plan |
Full command catalog: gald3r Wiki — Commands
Project Structure After Install
.gald3r/
├── TASKS.md ← master task index
├── BUGS.md ← bug index
├── PLAN.md ← milestones and strategy
├── PROJECT.md ← vision, mission, goals
├── CONSTRAINTS.md ← things the AI must never do
├── SUBSYSTEMS.md ← component registry
├── tasks/ ← individual task files (one per task)
├── bugs/ ← individual bug files
└── features/ ← PRD files
Contributing
See CONTRIBUTING.md. Issues and PRs welcome.
License
Fair Source License 1.1 (FSL-1.1-Apache) — see NOTICE for third-party attributions.
Yorumlar (0)
Yorum birakmak icin giris yap.
Yorum birakSonuc bulunamadi