Skills & Open Source Resources for AI Agents

/ Featured · Multi-Agent Orchestration

Skill delegate — Judgment vs. Execution Orchestration for AI Coding Agents

Orchestration pattern for AI coding agents: the frontier model preserves judgment and writes the Brief; cost-effective models execute. Works seamlessly on Claude Code, Antigravity, and ChatGPT/Codex with configurable model tiers.

Golden Rule: Your context is for judgment; subagent context is for material.
Three model tiers (judgment, worker, lookup) mapped across Claude Code, Antigravity, and Codex in an editable configuration block.
12-Point Engineering Brief template (Goal, Context, Decisions, Scope, Mirror, Traps, Acceptance, Return).
The 2 anti-blindness questions: Adjacent findings and Blast radius.
Per-writer isolation in Git worktrees (.worktrees/<plan>--<unit>) and tiered auditing (Tier 0/1/2).

Read guide & copy SKILL.md Raw file: /recursos/delegate/SKILL.md

delegate/SKILL.md · 12-Point Brief

## The Default Is to Delegate The burden of proof is on doing it yourself. Your context is for judgment; theirs is for material. ## The Brief (12 Differentiators) 1. Goal & Why 7. Invariants 2. Missing Context 8. Traps & Gotchas 3. Decisions Made 9. Scope Fence 4. Mirror This Path 10. Acceptance Cmds 5. Wrong Obvious Path 11. Adjacent Findings 6. Failure Policy 12. Blast Radius

/ Complete Catalog · 2 Resources Available

SKILL.MD · ORCHESTRATION

Skill delegate — Judgment vs. Execution Orchestration for AI Coding Agents

Orchestration pattern for AI coding agents: the frontier model preserves judgment and writes the Brief; cost-effective models execute. Works seamlessly on Claude Code, Antigravity, and ChatGPT/Codex with configurable model tiers.

Golden Rule: Your context is for judgment; subagent context is for material.
Three model tiers (judgment, worker, lookup) mapped across Claude Code, Antigravity, and Codex in an editable configuration block.

delegate/SKILL.md · 9 min read Explore & copy →

CASE STUDY · AI REDESIGN

Gran Café Tortoni Case Study: Redesigning an 1858 Institution with AI

Complete redesign of the iconic 1858 café: extracting design tokens from physical architecture, WCAG AA contrast compliance, AI image pipeline, interactive before/after splitters, and 8 production prompts.

Physical Design System: Canopy Green, Marble, and Stained Glass with 5.7:1 WCAG AA contrast ratio.
8 deterministic production prompts ready to copy with one-click button.

caso-tortoni · 8 prompts · 12 min read Read case study →

1-COMMAND INSTALLATION

Use them in Claude Code, Cursor, Antigravity, or your own repo

All skills are published under the MIT license. You can clone the complete bundle from GitHub or download individual SKILL.md files directly from their static URL.

View repository on GitHub Explore Imagine AI plans

terminal · bash

# Clone official skills repository git clone https://github.com/cgandini/agentic-skills.git # Install delegate in your agent cp -R agentic-skills/delegate ~/.claude/skills/ # Antigravity: cp -R agentic-skills/delegate ~/.gemini/config/skills/ # Codex: cp -R agentic-skills/delegate ~/.codex/skills/

/ Sovereign Software & AI Automation

Looking to implement this caliber of engineering in your business?

We design and build custom websites, Starter Kit operating portals, bespoke systems, and AI WhatsApp assistants at a one-time fixed price on your own infrastructure.

Tell us your project View transparent pricing