Galdr

Galdr

Slim

The open-source foundation

9 agents, 49 skills, 78 commands. File-first — no servers, no subscriptions. Works with Cursor, Claude, Codex, Gemini, and OpenCode.

Free · MITv1.2.1

Quick Start

git clone https://github.com/wrm3/galdr.git && cp -r galdr/template_full/.* your-project/
View on GitHub →

Features

What you get

TASKS.md + tasks/ lifecycle

Sequential task IDs with full state machine: pending → in-progress → awaiting-verification → complete.

Complexity scoring

Auto-score tasks 1–10. Score ≥ 7 triggers sub-task expansion.

Adversarial quality gates

Implementing agent marks [🔍]; separate session runs verification. Structurally prevents self-sign-off.

Bug tracking + BUGS.md

BUG-NNN format with severity, fix status, and discovery gate enforced by rules.

Sprint planning

ACTIVE_BACKLOG.md for prioritized sprint view. Dependency tracking with dependency graph.

CONSTRAINTS.md

Non-negotiable architectural guardrails that all agents must check before completing tasks.

IDEA_BOARD.md

Capture ideas without breaking flow. Promote to tasks when ready.

5-IDE parity

Same agents, skills, commands, and memory across Cursor, Claude Code, Gemini, Codex, and OpenCode.

Pre-commit hooks

Shell validation, user identity, INBOX check, and branch safety hooks.

Commands

@g-* command reference

CommandDescription
@g-goFull session orchestration: load context → implement → verify
@g-go-codeImplementation-only pass. Marks tasks [🔍], never [✅]
@g-go-reviewVerification pass — independent agent confirms [🔍] items
@g-statusSession context: active tasks, phase, goals, ideas
@g-planManage PLAN.md and feature staging lifecycle
@g-git-commitStructured commit with type prefix, task refs, agent footer
@g-git-pushSafety-gated push with branch checks
@g-git-sanityPre-push sanity check: lint, types, test smoke
@g-task-newCreate a new task with full spec
@g-task-updateUpdate task status, notes, or spec
@g-task-sync-checkValidate TASKS.md ↔ tasks/ files consistency
@g-bug-reportFile a new bug with classification + severity
@g-bug-fixFix workflow: branch → fix → verify → close
@g-code-reviewStructured code review: security, quality, performance
@g-groomingBacklog grooming: prioritize, split, estimate
@g-medkitHealth check + repair: migration, structure, sync
@g-cleanupRemove orphaned files and stale state
@g-setupInitialize galdr in a new project
@g-constraint-checkVerify implementation against active constraints
@g-constraint-addAdd a new architectural constraint
@g-idea-captureCapture idea to IDEA_BOARD.md
@g-idea-farmScan codebase for improvement opportunities
@g-idea-reviewReview and triage IDEA_BOARD.md
@g-learnStore and review project/global learned facts

Skills

Skill catalog

Core

g-skl-tasks

Own and manage all task data — TASKS.md, tasks/, status transitions, complexity scoring

g-skl-bugs

Bug tracking — BUGS.md, bugs/, quality metrics, discovery gates

g-skl-features

Feature lifecycle: staging → specced → committed → shipped

g-skl-plan

PLAN.md master strategy + feature staging

g-skl-project

PROJECT.md mission, goals, project linking

g-skl-constraints

CONSTRAINTS.md governance — active guardrails

g-skl-subsystems

SUBSYSTEMS.md registry + subsystem spec files

g-skl-git-commit

Structured commits with type prefix + task refs

g-skl-medkit

Project health check + migration + repair

g-skl-status

Session context display: tasks, phase, goals

g-skl-setup

Initialize galdr in a new project

g-skl-ideas

IDEA_BOARD.md — capture, review, promote

g-skl-review

Code review: security, quality, performance

g-skl-learn

File-only continual learning — agents self-report insights

g-skl-test

L1/L2/L3 test plans — fast, comprehensive, regression

g-skl-verify-ladder

Multi-level verification gates for autonomous completion

g-skl-dependency-graph

Task + subsystem dependency graphs

CLI

g-skl-cli-cursor

Cursor CLI: agent command, modes, Cloud Agent handoff

g-skl-cli-claude

Claude Code CLI: headless flags, session continuation, MCP

g-skl-cli-codex

OpenAI Codex CLI: approval modes, sandbox, provider flags

g-skl-cli-gemini

Gemini CLI: auth, checkpointing, config, extensions

Agents

Specialized agents

g-agnt-project

g-agnt-project.md

Manages PROJECT.md, goals, and project identity

g-agnt-task-manager

g-agnt-task-manager.md

Owns TASKS.md and individual task files

g-agnt-code-reviewer

g-agnt-code-reviewer.md

Adversarial code review — separate from implementing agent

g-agnt-qa-engineer

g-agnt-qa-engineer.md

Bug filing, QA workflow, quality metrics

g-agnt-verifier

g-agnt-verifier.md

Independent verification of [🔍] items

g-agnt-infrastructure

g-agnt-infrastructure.md

DevOps, CI/CD, Docker, cloud infra

g-agnt-ideas-goals

g-agnt-ideas-goals.md

IDEA_BOARD → task promotion + goal tracking

g-agnt-test

g-agnt-test.md

L1/L2/L3 test plan creation and execution

g-agnt-project-initializer

g-agnt-project-initializer.md

Spawns new galdr projects with full scaffold

Workflows

How it works

Typical workflow

  1. 1

    Create Task

    @g-task-new → task file with AC, subsystems, complexity

  2. 2

    Implement

    @g-go-code → agent implements, checks AC gate, marks [🔍]

  3. 3

    Verify

    @g-go-review → separate agent confirms criteria, marks [✅]

  4. 4

    Commit

    @g-git-commit → structured commit with task refs

  5. 5

    Review

    @g-code-review → security + quality gate before merge

Task Lifecycle

flowchart LR
  A([📋 pending]) --> B([🔄 in-progress])
  B --> C([🔍 awaiting-verification])
  C --> D([✅ complete])
  C --> E([❌ failed])
  E --> A
  style A fill:#1e293b,stroke:#94a3b8,color:#94a3b8
  style B fill:#1e293b,stroke:#f59e0b,color:#f59e0b
  style C fill:#1e293b,stroke:#60a5fa,color:#60a5fa
  style D fill:#1e293b,stroke:#4ade80,color:#4ade80
  style E fill:#1e293b,stroke:#f87171,color:#f87171

Mermaid diagram — paste into any Mermaid renderer to visualize