liza-masCommunity·11 Agent Skills Included

liza

Hardened multi-agent coding with enforced reviews and specs

Coordinates doer and reviewer agents through spec-driven coding sprints with enforced state machines and isolated git worktrees. Eliminates fake progress, review churn, and untraceable decisions via adversarial review loops, checkpoint gates, and auditable blackboards. Includes skills for spec writing, code review, debugging, red testing, architecture planning, and log analysis to ship reliable code faster.

346|50

↗

npx skills add liza-mas/liza --all -g -y

All Skills in This Repository (11)

Pure Emerald Level Indicators

📦 In Repo

liza-mas

software-architecture-review

Analyze software architecture through a three-phase Discovery, Analysis, and Recommendations process.

code-review

Guide engineers through structured code reviews with risk-based modes and checklists.

debugging

Guide teams through structured debugging protocols to identify root causes and verify fixes.

systemic-thinking

Analyze systemic coherence and cascading risks across design artifacts.

code-cleaning

Clean and refactor staged Python code during pre-commit flows.

feynman

Explain complex topics using the Feynman technique with analogies and refinement.

testing

Design deterministic tests using GIVEN/WHEN/THEN structure.

adr-backfill

Reconstruct MADR-formatted Architecture Decision Records from commits and documentation.

generic-subagent

Delegate read-only research tasks to subagents and return structured digests.

spec-review

Analyze specification documents and generate a structured issue log.

spec-backfill

Analyze code, docs, and git history to backfill missing specifications.

Community

Intermediate

Frequently Asked Questions

FAQPage Schema

How to install Liza?▼

Run `npx skills add liza-mas/liza --all -g -y` in your terminal to install all skills in this suite globally.

What problem does Liza solve?▼

Liza stops coding agents from faking progress by enforcing a behavioral contract, dual doer/reviewer reviews, and deterministic state machines. Every task is tracked on an auditable blackboard with human checkpoints between sprints.

Which AI coding tools does Liza support?▼

Liza wraps provider CLIs like Claude Code, Codex CLI, OpenCode, Kimi, Mistral, and Gemini, so your existing subscriptions work without API keys or per-token billing.

How does Liza keep multi-agent code reliable?▼

A Go supervisor enforces worktree isolation, TDD gates, and 43+ task validation rules while agents handle judgment. Circuit breakers, crash recovery, and log analysis skills catch loops and regressions automatically.

Can non-developers use Liza?▼

Yes. You steer between sprints by reviewing plain-language checkpoint summaries, while agents handle specs, code, and tests autonomously within each sprint.