skills.rest
Shubh2310-developer
Shubh2310-developerCOMMUNITY
Explore all skills by this author →
GitHub

loki-mode

Orchestrates multi-agent swarms to build, test, and deploy products from a PRD.

2|Updated Jan 10, 2026
Search Tags:#deployment#claude code#multi-agent#orchestration#prd#autonomous#task queue
One-click install
npx skills add https://github.com/Shubh2310-developer/ENGUNITYCORE --skill loki-mode-shubh2310-developer
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: loki-mode
Source: https://github.com/Shubh2310-developer/ENGUNITYCORE/tree/main/.claude/skills/loki-mode
Command: npx skills add https://github.com/Shubh2310-developer/ENGUNITYCORE --skill loki-mode-shubh2310-developer

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve? Coordinating dozens of AI agents across engineering, QA, DevOps, and business workflows requires manual orchestration, state tracking, and failure recovery that single-agent coding sessions cannot provide. ## Core Features & Use Cases - Multi-Agent Orchestration: Dispatches 37 specialized agent types across 7 swarms (Engineering, Operations, Business, Data, Product, Growth, Review) using the Task tool with model-tiered routing (Opus for planning, Sonnet for development, Haiku for tests). - RARV Self-Verification Cycle: Every iteration follows Reason-Act-Reflect-Verify with git checkpoints, continuity logs, and automatic retry on test failure. - Quality Gates & Recovery: Enforces blind 3-reviewer parallel code review, anti-sycophancy devil's advocate checks, circuit breakers, dead-letter queues, and exponential-backoff resume on rate limits. - Use Case: Provide a PRD for a full-stack web app and let the system autonomously progress through discovery, architecture, development, QA, deployment, and growth phases while you monitor a live dashboard. ## Quick Start Run ./autonomy/run.sh with the path to your PRD file after launching Claude Code with the --dangerously-skip-permissions flag.

Frequently Asked Questions about loki-mode

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I run Loki Mode with a PRD file?▼

Run ./autonomy/run.sh followed by the path to your PRD markdown file from the skill directory. The script checks prerequisites, initializes the .loki/ state directory, starts a status monitor, and launches Claude Code with auto-resume on rate limits.

How does multi-agent orchestration work in Claude Code?▼

The orchestrator dispatches subagents via the Task tool, selecting models by task type: Opus for planning and architecture, Sonnet for development and integration tests, Haiku for unit tests and monitoring. Simple tasks route directly to specialists while complex tasks use supervisor coordination.

Does Loki Mode require special permissions to run?▼

Yes, it requires launching Claude Code with the --dangerously-skip-permissions flag so agents can execute commands autonomously. It also needs Python 3, Git, and optionally cloud provider credentials for the deployment phase.

What happens when Claude hits a rate limit during autonomous runs?▼

The autonomy runner saves a state checkpoint, waits with exponential backoff starting at 60 seconds up to a 1-hour cap, then resumes from the exact interruption point. It retries up to 50 times by default, configurable via LOKI_MAX_RETRIES.

Why does editing run.sh crash an active Loki Mode session?▼

Bash reads scripts incrementally during execution, so modifying autonomy/run.sh while it runs corrupts the executing process. Document fixes in .loki/CONTINUITY.md under Pending Fixes and apply them after the session ends.

When should I not use Loki Mode?▼

Skip it when you need human approval before actions, want to review a plan first, or have a single small task. The system is designed for zero-intervention autonomous execution and never asks questions or waits for confirmation.

Related Skills

Explore More Software Engineering Skills→
firdaus12pfirdaus12p
8

brainstorm-architecture

Facilitate structured system architecture discussions covering tech stack, API design, and ADR documentation.

Community
Advanced
pe-menezespe-menezes

experiment-design

Guide product experiment design, execution, and analysis for A/B testing.

Community
Advanced
librefanglibrefang
10

gcp

Automate GCP infrastructure design, deployment, and governance with IAM and Terraform.

Official
Advanced
DagimAlemayehuuDagimAlemayehuu

agentic-engineering-governance

Define authority, risk, budget, and approval boundaries for agent-driven projects.

Community
Intermediate