Coordinate Zoo 2.1 workflows with delegated research, review, and browser-verification subagents.
npx skills add https://github.com/andreyvit/zoo --skill zoo-lite
Please help me install this Agent Skill. Skill: zoo-lite Source: https://github.com/andreyvit/zoo/tree/main/.codex/skills/zoo-lite Command: npx skills add https://github.com/andreyvit/zoo --skill zoo-lite
Zoo Lite provides a structured, plan-driven framework to execute complex tasks by coordinating a top-level orchestrator with research, browser verification, and review subagents, ensuring reproducible plans and traceable results.
Start a Zoo Lite session and run zoo-plan to define the first (next) subtask
High-intent search queries and answers about installing and using this skill.
Spec-driven workflow orchestration coordinates a top-level planner with delegated research, review, and browser-verification subagents. It executes formal planning cycles and uses evidence-backed validation to ensure reproducible plans and traceable results across implementation and reviews.
A spec-based task lifecycle uses a Bureau MCP-driven process to maintain a single active next subtask at any time. It enforces a cadence of plan, review, and uberreview phases, ensuring complex tasks progress sequentially with structured evidence management and durable documentation.
Start a workflow session by initiating the orchestrator and running the planning command to define the first next subtask. This generates the initial formal plan, establishing the structured execution loop for subsequent research, implementation, and browser-verification phases.
Browser-impact verification is supported through dedicated browser_verifier subagents. These agents execute checks during the workflow lifecycle and generate mandatory evidence reports, ensuring that implementation changes are validated against actual browser behavior.
Use a spec-driven orchestrator for complex tasks requiring formal planning cycles, bounded research questions, and multi-phase validation. Standard agents lack the structured plan, review, and uberreview cadences needed for reproducible plans and evidence-backed execution across multiple subagents.
Plan-driven workflows with delegated subagents require mandatory frontmatter in SKILL.md and impose strict plan/review cadences. This overhead makes them less suitable for simple, single-step tasks where lightweight execution is preferred over formal planning cycles and evidence management.
pradeepmouliGenerate baseline-spec.md and current-state.md from repository analysis.
benjaminshoemakerAnalyze repository plans and compute completion metrics from Markdown checkboxes.
ryan-mtOrchestrate an 8-phase research pipeline with multi-source verification and synthesis.
zapabobOrchestrate multi-agent workflows with task decomposition, handoffs, and guardrails.
