Skip to content
/
Tier B A11 Governance

awesome-gemini-cli-subagents

ankitmundada-gemini-subagents · ankitmundada/awesome-gemini-cli-subagents · ★ 33 · last commit 2026-05-10

Primitive shape 128 total
Subagents 128
00

Summary

awesome-gemini-cli-subagents — Summary

awesome-gemini-cli-subagents is a curated collection of 128+ persona-MD subagent definitions for Google's Gemini CLI, organized across 10 thematic categories (core development, language specialists, infrastructure, quality/security, data/AI, developer experience, specialized domains, business/product, meta-orchestration, research/analysis). Each agent is a markdown file with a YAML frontmatter block that specifies the agent's name, description, tool permissions, model assignment (Gemini model routing), temperature, and max turn count, followed by a freeform system prompt with checklists, communication protocols, and structured workflow phases. The collection ships an interactive bash installer (install-agents.sh) that lets users browse categories, select agents, and copy them to ~/.gemini/agents/ or .gemini/agents/. A meta agent-installer agent within the collection can itself be used to install other agents from inside Gemini CLI. The project's distinctive engineering choice is per-agent model routing: each .md file declares which Gemini model to use (gemini-3-pro-preview for deep reasoning, gemini-3-flash-preview for quick tasks), making the collection the gemini-cli analog to BMAD-METHOD's persona packs but with explicit multi-model dispatch built into every agent definition.

differs_from_seeds: Most similar to BMAD-METHOD (Archetype 1 — persona-MD files defining specialized roles), but differs in three ways: it targets Gemini CLI rather than Claude Code; each agent carries explicit model: and temperature: frontmatter enabling automatic multi-model routing absent from BMAD; and the collection is domain-organized (128 specialists) rather than SDLC-phase-organized (6 BMAD personas). Unlike superpowers or spec-driver, there are no hooks, no spec files, and no lifecycle enforcement — agents are invoked on-demand via Gemini CLI's /agents interface.

01

Overview

awesome-gemini-cli-subagents — Overview

Origin

Created by Ankit Mundada (ankitmundada, GitHub), first published February 2026. The project was created specifically to exploit Gemini CLI's experimental subagent feature (experimental.enableAgents: true in Gemini CLI settings.json).

Philosophy

The project README states:

"This repository serves as the definitive collection of Gemini CLI subagents, specialized AI assistants designed for specific development tasks."

The driving belief is that domain specialization via persona injection dramatically improves output quality for specific tasks. From the README's explanation of subagents:

"Subagents are specialized AI assistants that enhance Gemini CLI's capabilities by providing task-specific expertise. They act as dedicated helpers that Gemini CLI can call upon when encountering particular types of work."

Key design principles surfaced in the project:

  1. Independent context windows: Each subagent "operates within its own isolated context space, preventing cross-contamination between different tasks."
  2. Domain-specific intelligence: "Specialized prompts and configurations lead to better results in specific domains."
  3. Granular tool permissions: Read-only agents get fewer tools; code-writing agents get full tool access.
  4. Smart model routing: Temperature and model are chosen per agent role, not globally.

Warning in README

"Sub-agents in Gemini CLI are currently experimental. You must enable them in your settings.json... Sub-agents currently operate in 'YOLO mode', meaning they may execute tools without individual user confirmation for each step."

The "YOLO mode" caveat is prominently displayed as an experimental warning.

Temperature Optimization

The project documents a temperature optimization strategy:

  • Code generation: 0.1 (deterministic)
  • Debugging: 0.05 (most precise)
  • Architecture: 0.35 (balanced exploration)
  • Research: 0.45 (exploratory)

This is encoded in a TEMPERATURE_OPTIMIZATION_SUMMARY.md file referenced in the README.

Gemini CLI Integration

Agents are stored in:

  • ~/.gemini/agents/ — global (all projects)
  • .gemini/agents/ — project-specific (overrides global on name conflict)

Gemini CLI auto-selects agents based on their description field, or users can explicitly invoke: "Have the code-reviewer subagent analyze my latest commits".

02

Architecture

awesome-gemini-cli-subagents — Architecture

Distribution

  • Standalone repository: Clone or use installer script
  • Manual install: Copy .md files from categories/ to ~/.gemini/agents/ or .gemini/agents/
  • Interactive installer: ./install-agents.sh (browse + select via interactive menu)
  • Standalone installer: curl -sO .../install-agents.sh && chmod +x && ./install-agents.sh (fetches from GitHub API, no clone required)
  • Meta-agent install: Copy agent-installer.md to ~/.gemini/agents/, then ask Gemini CLI to install others

Required Runtime

  • Gemini CLI (Google's official CLI)
  • Gemini CLI settings.json must have "experimental": {"enableAgents": true}
  • curl (for standalone installer)
  • No Node.js, Python, or other runtimes required for the agents themselves

Repository Tree

awesome-gemini-cli-subagents/
├── README.md
├── GEMINI.md               # Gemini CLI guidance for this repo
├── CONTRIBUTING.md
├── install-agents.sh       # Interactive bash installer
├── categories/
│   ├── 01-core-development/     # 11 agents
│   ├── 02-language-specialists/ # 26 agents
│   ├── 03-infrastructure/       # 14 agents
│   ├── 04-quality-security/     # 14 agents
│   ├── 05-data-ai/              # 12 agents
│   ├── 06-developer-experience/ # 14 agents
│   ├── 07-specialized-domains/  # 12 agents
│   ├── 08-business-product/     # 11 agents
│   ├── 09-meta-orchestration/   # 10 agents (+ 1 external link)
│   └── 10-research-analysis/    # 6 agents
└── TEMPERATURE_OPTIMIZATION_SUMMARY.md

Agent File Format

Each agent is a Markdown file with YAML frontmatter:

---
name: agent-name
description: "Natural language trigger description for auto-selection"
kind: local
tools:
  - read_file
  - write_file
  - run_shell_command
  - search_file_content
model: gemini-3-pro-preview   # or gemini-3-flash-preview
temperature: 0.1              # 0.05 to 0.55
max_turns: 20                 # typically 20-30
---

[System prompt content]

Target AI Tool

Google Gemini CLI exclusively. The agent format uses Gemini CLI's proprietary kind: local subagent spec. Not compatible with Claude Code, Cursor, or other MCP clients.

Install Configuration

After installation, agents are activated via Gemini CLI's /agents command or referenced by name in conversation.

03

Components

awesome-gemini-cli-subagents — Components

Subagents by Category (128+ total)

01 — Core Development (11 agents)

api-designer, backend-developer, electron-pro, frontend-developer, fullstack-developer, graphql-architect, microservices-architect, mobile-developer, ui-designer, websocket-engineer, wordpress-master

02 — Language Specialists (26 agents)

typescript-pro, sql-pro, swift-expert, vue-expert, angular-architect, cpp-pro, csharp-developer, django-developer, dotnet-core-expert, dotnet-framework-4.8-expert, elixir-expert, flutter-expert, golang-pro, java-architect, javascript-pro, powershell-5.1-expert, powershell-7-expert, kotlin-specialist, laravel-specialist, nextjs-developer, php-pro, python-pro, rails-expert, react-specialist, rust-engineer, spring-boot-engineer

03 — Infrastructure (14 agents)

azure-infra-engineer, cloud-architect, database-administrator, deployment-engineer, devops-engineer, devops-incident-responder, incident-responder, kubernetes-specialist, network-engineer, platform-engineer, security-engineer, sre-engineer, terraform-engineer, windows-infra-admin

04 — Quality & Security (14 agents)

accessibility-tester, ad-security-reviewer, architect-reviewer, chaos-engineer, code-reviewer, compliance-auditor, debugger, error-detective, penetration-tester, performance-engineer, powershell-security-hardening, qa-expert, security-auditor, test-automator

05 — Data & AI (12 agents)

ai-engineer, data-analyst, data-engineer, data-scientist, database-optimizer, llm-architect, machine-learning-engineer, ml-engineer, mlops-engineer, nlp-engineer, postgres-pro, prompt-engineer

06 — Developer Experience (14 agents)

build-engineer, cli-developer, dependency-manager, documentation-engineer, dx-optimizer, git-workflow-manager, legacy-modernizer, mcp-developer, powershell-ui-architect, powershell-module-architect, refactoring-specialist, slack-expert, tooling-engineer, visual-asset-generator

07 — Specialized Domains (12 agents)

api-documenter, blockchain-developer, embedded-systems, fintech-engineer, game-developer, iot-engineer, m365-admin, mobile-app-developer, payment-integration, quant-analyst, risk-manager, seo-specialist

08 — Business & Product (11 agents)

business-analyst, content-marketer, customer-success-manager, legal-advisor, product-manager, project-manager, sales-engineer, scrum-master, technical-writer, ux-researcher, ai-content-editor

09 — Meta & Orchestration (10 local + 1 external)

agent-installer, agent-organizer, context-manager, error-coordinator, it-ops-orchestrator, knowledge-synthesizer, multi-agent-coordinator, performance-monitor, task-distributor, workflow-orchestrator, pied-piper (external, https://github.com/sathish316/pied-piper/)

10 — Research & Analysis (6 agents)

research-analyst, search-specialist, trend-analyst, competitive-analyst, market-researcher, data-researcher

Scripts (1)

Script Purpose
install-agents.sh Interactive bash script to browse categories, select agents, install to global/local .gemini/agents/ — supports local-clone and remote-GitHub-API modes

Templates (0)

No template files separate from the agent definitions themselves.

Commands (0)

No slash-commands, no CLI subcommands beyond what install-agents.sh provides.

Hooks (0)

No lifecycle hooks. Agent auto-selection is handled by Gemini CLI based on the description field in frontmatter.

Tool Permissions Per Agent Class

Agent class tools assigned
Read-only (reviewers, auditors) read_file, search_file_content
Research agents read_file, search_file_content, web_fetch, web_search
Code writers read_file, write_file, run_shell_command, search_file_content
Documentation agents read_file, write_file, search_file_content, web_fetch, web_search
05

Prompts

awesome-gemini-cli-subagents — Prompts

All prompts are the body text of .md agent files (below the YAML frontmatter). Two representative verbatim excerpts follow.

Prompt 1 — api-designer (categories/01-core-development/api-designer.md)

Frontmatter:

name: api-designer
description: API architecture expert designing scalable, developer-friendly interfaces.
kind: local
tools: [read_file, write_file, run_shell_command, search_file_content]
model: gemini-3-pro-preview
temperature: 0.35
max_turns: 25

System prompt opening (verbatim):

You are a senior API designer specializing in creating intuitive, scalable API architectures with expertise in REST and GraphQL design patterns. Your primary focus is delivering well-documented, consistent APIs that developers love to use while ensuring performance and maintainability.

When invoked:
1. Read relevant files for existing API patterns and conventions
2. Review business domain models and relationships
3. Analyze client requirements and use cases
4. Design following API-first principles and standards

Checklist section (verbatim):

API design checklist:
- RESTful principles properly applied
- OpenAPI 3.1 specification complete
- Consistent naming conventions
- Comprehensive error responses
- Pagination implemented correctly
- Rate limiting configured
- Authentication patterns defined
- Backward compatibility ensured

Prompting technique: Persona injection ("You are a senior API designer...") + ordered invocation steps + domain checklist. The checklist pattern is the primary behavioral guardrail. The communication protocol section uses structured JSON as a convention for inter-agent messaging.

Prompt 2 — multi-agent-coordinator (categories/09-meta-orchestration/multi-agent-coordinator.md)

Frontmatter:

name: multi-agent-coordinator
description: Expert multi-agent coordinator specializing in complex workflow orchestration...
model: gemini-3-pro-preview
temperature: 0.4
max_turns: 30

System prompt opening (verbatim):

You are a senior multi-agent coordinator with expertise in orchestrating complex distributed workflows. Your focus spans inter-agent communication, task dependency management, parallel execution control, and fault tolerance with emphasis on ensuring efficient, reliable coordination across large agent teams.

Progress reporting section (verbatim):

Progress tracking:
{
  "agent": "multi-agent-coordinator",
  "status": "coordinating",
  "progress": {
    "active_agents": 87,
    "messages_processed": "234K/min",
    "workflow_completion": "94%",
    "coordination_efficiency": "96%"
  }
}

Delivery notification (verbatim):

"Multi-agent coordination completed. Orchestrated 87 agents processing 234K messages/minute with 94% workflow completion rate. Achieved 96% coordination efficiency with zero deadlocks and 99.9% message delivery guarantee."

Prompting technique: Role injection + enumerated domain checklists + structured JSON formatting conventions for reporting. The "coordination excellence checklist" pattern encodes measurable KPIs (coordination overhead < 5%, deadlock prevention 100%) directly into the system prompt as behavioral expectations. The delivery notification is a prescribed output format template.

Key Observations

  1. Uniform structure: All 128+ agents follow the same template: role-intro paragraph + "when invoked" steps + domain checklists + communication protocol + workflow phases + integration-with-other-agents list.
  2. Temperature as quality signal: Lower temperature (0.05) for debuggers, higher (0.45) for research — this is explicitly documented and encoded in frontmatter.
  3. Model routing via frontmatter: The model: field routes to gemini-3-pro-preview (deep reasoning) or gemini-3-flash-preview (quick tasks) — unique in this corpus.
  4. Inter-agent JSON protocol: All agents include a JSON "communication protocol" section defining how to request context from other agents. This is a convention enforced only by the agent's system prompt, not by Gemini CLI infrastructure.
09

Uniqueness

awesome-gemini-cli-subagents — Uniqueness

differs_from_seeds

Most similar to BMAD-METHOD in being a collection of persona-MD files defining specialized roles (both are Archetype 1 — skills-only behavioral frameworks). But three architectural deltas separate them: (1) This collection targets Gemini CLI rather than Claude Code — the frontmatter format (kind: local, model:, temperature:, max_turns:) is Gemini-specific and incompatible with Claude Code's skill format. (2) Every agent carries explicit model: and temperature: frontmatter enabling automatic multi-model routing — BMAD has no model routing; all BMAD agents use whatever model the Claude Code session uses. (3) The collection is domain-organized (128 specialist personas across 10 categories) rather than SDLC-phase-organized (BMAD's 6 workflow-step personas). Compared to superpowers or spec-driver, this collection has no hooks, no spec files, no lifecycle enforcement, and no TDD guardrails — it is a pure role-injection library.

Positioning

The collection fills the "BMAD for Gemini CLI" niche. Its competitive advantage is breadth (128 specialists covering domains from blockchain to Windows infrastructure) and the model-routing embedded in each agent. It does not attempt to be a full development methodology — it is a specialist library users bolt onto Gemini CLI's existing conversational workflow.

What Makes It Distinct

  1. Per-agent model routing: No other collection in this corpus embeds explicit model selection in each agent definition. The model: frontmatter field makes model routing declarative and per-specialization rather than global.

  2. Temperature optimization as first-class concern: The TEMPERATURE_OPTIMIZATION_SUMMARY.md document and per-agent temperature values (0.05 for debuggers, 0.35 for architects, 0.45 for researchers) represent unusual deliberateness about inference settings.

  3. Meta-agent bootstrapping: The agent-installer agent creates a self-referential install experience — you install one agent that can then install all others from inside Gemini CLI, without needing a terminal shell.

  4. "YOLO mode" explicit acknowledgment: The README's prominent warning about "YOLO mode" (tools execute without per-step confirmation) is an unusually honest safety disclosure not seen in most collections in this corpus.

  5. Gemini-CLI exclusivity: Unlike most collections which target Claude Code (or claim multi-tool portability), this is the only collection in the corpus explicitly designed for Google Gemini CLI.

Observable Failure Modes

  • Experimental feature dependency: The entire collection depends on experimental.enableAgents: true in Gemini CLI. If Google deprecates or changes this feature, all agents become inoperative.
  • YOLO mode risk: Agents with run_shell_command access operate without per-step confirmation. Poorly scoped invocations can make unintended file changes or run destructive commands.
  • Aspirational orchestration: Category 09 describes hierarchical multi-agent workflows with 87 concurrent agents, but Gemini CLI's experimental subagent infrastructure may not support true parallel agent execution — the coordination patterns are described in system prompts, not enforced by the platform.
  • Naming conflicts: Project-local agents silently override global agents with the same name — documented, but easy to misconfigure.
  • No validation: There is no CI/CD to validate agent frontmatter or system prompt quality. Contributed agents may have inconsistent tool assignments or broken JSON in communication protocol blocks.

Explicit Antipatterns

None documented. The README focuses on capabilities and installation, not what to avoid.

04

Workflow

awesome-gemini-cli-subagents — Workflow

Overview

There is no prescribed multi-phase development workflow. The collection is a library of specialist agents, not a pipeline. Each agent is self-contained with its own internal workflow phases, but the collection imposes no cross-agent sequencing.

Installation Workflow

Phase Action Artifact
1. Install Gemini CLI User prerequisite Gemini CLI configured
2. Enable agents Add experimental.enableAgents: true to settings.json Enabled agent feature
3. Select agents Run ./install-agents.sh or copy files manually Agent .md files
4. Deploy Files placed in ~/.gemini/agents/ or .gemini/agents/ Installed agents
5. Invoke Use /agents in Gemini CLI or ask Gemini to use a specific agent Agent response

Per-Agent Workflow Structure

Each agent's system prompt defines its own 3-phase workflow:

  1. Context Assessment — Agent requests relevant files/context via a structured JSON message
  2. Implementation Phase — Agent executes its domain task
  3. Delivery — Agent emits a completion report with metrics

Example from api-designer:

Phase 1: Domain Analysis → business requirements, data models
Phase 2: API Specification → endpoint design, schemas, auth
Phase 3: Developer Experience → documentation, SDKs, examples

Approval Gates

None at the collection level. Individual agents operate in Gemini CLI's "YOLO mode" (no per-step confirmation by default). The README explicitly warns:

"Sub-agents currently operate in 'YOLO mode', meaning they may execute tools without individual user confirmation for each step."

Spec Format

None. Agents do not produce or consume spec files. Each agent is invoked conversationally or via explicit mention.

Inter-Agent Communication Pattern

Orchestration agents (category 09) define a JSON message protocol for requesting context from other agents:

{
  "requesting_agent": "api-designer",
  "request_type": "get_api_context",
  "payload": {
    "query": "API design context required: ..."
  }
}

This is a soft convention documented in agent system prompts, not enforced by Gemini CLI itself.

Progress Reporting Pattern

Orchestration agents include progress tracking JSON blocks in their system prompts as formatting conventions:

{
  "agent": "multi-agent-coordinator",
  "status": "coordinating",
  "progress": {
    "active_agents": 87,
    "workflow_completion": "94%"
  }
}
06

Memory Context

awesome-gemini-cli-subagents — Memory & Context

State Storage

None at the collection level. Memory is managed by Gemini CLI's native context window per agent invocation. Agents have independent context windows from the main conversation thread (this is a Gemini CLI platform feature, not something the collection implements).

Persistence

No cross-session persistence. Each agent invocation starts fresh. The collection ships no database, no file-based memory, no state files.

Context Compaction

Not applicable at the collection level. Gemini CLI manages the context window for each agent. Agents are not configured with compaction strategies.

Cross-Session Handoff

None. Individual agents document their workflows in their system prompts but accumulate no persistent state.

Memory Type

None (from the collection's perspective). Gemini CLI provides isolated in-session context per agent.

State Files

No state files written by the collection. The only files the collection manages are:

  • ~/.gemini/agents/*.md — agent definition files (installed by the user)
  • .gemini/agents/*.md — project-specific overrides

Context Window Isolation

The README explicitly calls out "independent context windows" as a key feature:

"Every subagent operates within its own isolated context space, preventing cross-contamination between different tasks and maintaining clarity in the primary conversation thread."

This isolation is a Gemini CLI platform capability. The collection leverages it by defining max_turns per agent (typically 20-30) to bound context growth.

max_turns as Implicit Context Management

Each agent's frontmatter includes max_turns: 20-30. This is the only context management lever the collection controls — it limits how many turns an agent can take before the conversation must return to the main thread.

07

Orchestration

awesome-gemini-cli-subagents — Orchestration

Multi-Agent

Yes — explicitly. Category 09 (meta-orchestration) contains 10 agents designed to coordinate others: multi-agent-coordinator, workflow-orchestrator, agent-organizer, task-distributor, error-coordinator, context-manager, performance-monitor, knowledge-synthesizer, it-ops-orchestrator, agent-installer.

Orchestration Pattern

Hierarchical (coordinator + specialists). The multi-agent-coordinator and workflow-orchestrator agents describe a master-worker pattern where one orchestrating agent assigns tasks to specialists. The system prompt of multi-agent-coordinator enumerates supported coordination patterns: master-worker, peer-to-peer, hierarchical, publish-subscribe, request-reply, pipeline, scatter-gather, consensus-based.

However, the actual enforcement is prompt-level only — Gemini CLI does not natively support programmatic agent spawning. The patterns described are aspirational conventions the agents' system prompts encode, not hard-wired runtime behavior.

Max Concurrent Agents

Unknown — determined by Gemini CLI's platform limits for the experimental subagent feature, not by this collection.

Isolation Mechanism

Gemini CLI platform-provided isolated context windows per agent. No git worktrees, containers, or process sandboxing.

Consensus Mechanism

None enforced. The multi-agent-coordinator system prompt references "consensus-based" as a pattern option but no actual consensus protocol is implemented.

Prompt Chaining Pattern

Yes — documented in agent system prompts. The workflow-orchestrator and task-distributor agents define workflows where one agent's output feeds another agent's input as context. This is a soft convention via the inter-agent JSON communication protocol embedded in prompts.

Multi-Model Usage

Yes — this is the collection's primary technical distinction. Each agent's frontmatter carries a model: field:

Model Use case Example agents
gemini-3-pro-preview Deep reasoning — architecture, security, financial logic security-auditor, architect-reviewer, fintech-engineer
gemini-3-pro-preview Everyday coding python-pro, backend-developer, devops-engineer
gemini-3-flash-preview Quick tasks — docs, search, dependency checks documentation-engineer, seo-specialist, build-engineer

Users can override any agent's model by editing the model field or setting model: inherit to use the main conversation model.

Execution Mode

Interactive-loop (on-demand invocation via Gemini CLI conversation or /agents command). No background daemons, no scheduled execution.

Crash Recovery

None. No retry logic, no state checkpoints.

Context Compaction Handling

No. Governed by Gemini CLI platform, not by this collection.

Streaming Output

Determined by Gemini CLI platform. Not configured by the collection.

08

Ui Cli Surface

awesome-gemini-cli-subagents — UI & CLI Surface

Dedicated CLI Binary

None for the collection itself. The agents are used through Gemini CLI's own interface.

Installer Script

install-agents.sh is a bash script (not a binary) with an interactive terminal menu:

Feature Details
Mode Interactive bash TUI (color-coded menus)
Scope selection Global (~/.gemini/agents/) or project-local (.gemini/agents/)
Source mode Local clone or remote GitHub API
Browse Navigate 10 categories, list agents, select individually or all
Install/uninstall Copy or delete .md files
Dependencies curl for remote mode

Example usage:

./install-agents.sh
# → Select: Global or Project-local installation
# → Select: Local (clone present) or Remote (GitHub API)
# → Browse categories, select agents
# → Agents copied to target directory

IDE Integration

None. The collection targets Gemini CLI, which is a terminal application. No VS Code extension, no IDE plugin.

Gemini CLI Integration Surface

Once agents are installed, users interact via:

  • /agents command in Gemini CLI — lists and manages installed agents
  • Explicit mention in conversation: "Have the code-reviewer subagent analyze my latest commits"
  • Auto-selection: Gemini CLI reads the description field of each agent and may auto-invoke the best-matching agent based on conversation context

Meta-Agent Installer

A unique surface: agent-installer.md (in category 09) can itself be installed as an agent, then used within Gemini CLI to browse and install other agents from this repository:

# Install just the installer agent
curl -s https://raw.githubusercontent.com/ankitmundada/awesome-gemini-cli-subagents/main/categories/09-meta-orchestration/agent-installer.md -o ~/.gemini/agents/agent-installer.md

# Then inside Gemini CLI:
"Use the agent-installer to show me available categories"
"Find PHP agents and install php-pro globally"

Observability

None beyond what Gemini CLI provides natively. The collection ships no logging, no metrics, no audit trail.

Related frameworks

same archetype · same primary tool · same memory type

OpenHarness ★ 13k
A11 Governance

Open-source Python agent runtime providing complete harness infrastructure: tools, memory, governance, swarm coordination, and…

Trae Agent ★ 12k
A11 Governance

Research-friendly open-source CLI coding agent by ByteDance, designed for academic ablation studies and modular LLM provider…

Sweep AI ★ 7.7k
A11 Governance

Autonomous GitHub bot that converts issues to pull requests using a sequential multi-agent pipeline.

Agent Governance Toolkit (microsoft) ★ 2.3k
A11 Governance

Enterprise-grade AI agent governance: YAML policy enforcement, 12-vector prompt injection defense, zero-trust identity,…

TDD Guard ★ 2.1k
A11 Governance

Mechanically enforces the Red-Green-Refactor TDD cycle by blocking file writes that violate TDD principles via a PreToolUse hook…

Agentic Coding Flywheel Setup (ACFS) ★ 1.5k
A11 Governance

Take a complete beginner from laptop to three AI coding agents running on a VPS in 30 minutes via an idempotent manifest-driven…