Skills DirectorySkills Directory
SkillsLearnSecurityCategoriesDocsCommunityBlog
Sign InSubmit Skill
Skills Directory

Security-tested agent skills for Claude, coding agents, and AI workflows.

Directory

  • Browse Skills
  • All Skills A–Z
  • Claude Skills
  • Claude Code Skills
  • Agent Skills
  • Categories
  • Submit a Skill

Learn

  • Learn Hub
  • Install Claude Skills
  • Write SKILL.md
  • Skills vs MCP
  • Directories Compared

Security

  • Security
  • Methodology
  • Secure Claude Skills
  • Security Badges

Company

  • About
  • Community
  • Blog
  • API Docs
  • Advertise

2026 Skills Directory. All rights reserved.

Back to skills

Outcome Rubric Verification

ASecurity

Use when implementing rubric-driven agent iteration with independent verifier — clean context, hill-climbing, choose-not-to-show. Triggers on "outcome rubric", "rubric verification", "independent verifier", "hill climbing rubric", "choose not to show".

2 stars
0 votes
0 copies
0 views
Added 9/19/2026
ai-agentsgogitapi

Works with

cliapi

Security Analysis

A100/100

Scanned 9/19/2026

Install to Claude Code

$npx -y skills add majinmagros/magros.ai-skills --skill outcome-rubric-verification --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Outcome Rubric Verification?

Add the live security badge to your README — it updates automatically with every re-scan.

Security grade badge for Outcome Rubric Verification
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/majinmagros-outcome-rubric-verification/badge)](https://www.skillsdirectory.com/skills/majinmagros-outcome-rubric-verification)

More formats (shields.io, HTML) on the badges page.

Download Zip
Files
SKILL.md
---
name: outcome-rubric-verification
description: Use when implementing rubric-driven agent iteration with independent verifier — clean context, hill-climbing, choose-not-to-show. Triggers on "outcome rubric", "rubric verification", "independent verifier", "hill climbing rubric", "choose not to show".
metadata:
  origin: ECC
  source_docs:
    - https://docs.anthropic.com/en/docs/managed-agents/outcomes
    - https://github.com/anthropics/managed-agents-sdk
  video_source: "hm8NzEd5io0 - How founders build on Claude Managed Agents (Claude Oficial)"
  related_skills:
    - loop-design-check
    - grill-with-docs
    - santa-method
    - verification-loop
    - gan-style-harness
---

# Skill: outcome-rubric-verification — Verificação por Rubrica com Verifier Independente

Padrão do **Managed Agents Outcomes API**: iteração dirigida por rubrica com **verifier independente em context window limpo**, hill-climbing até satisfazer a rubrica, e **choose-not-to-show** se falhar. Implementação completa em `references/implementation.md`.

## Quando usar

- Você precisa de **verificação rigorosa** de output de agent antes de mostrar ao usuário
- Quer separar **geração** de **validação** (clean context window)
- Precisa de **hill-climbing automático** contra critérios definidos
- Quer implementar **choose-not-to-show** em vez de mostrar resultado ruim
- Está construindo **briefs, relatórios, código, análises** que precisam ser corretos

## Quando NÃO usar

- Verificação determinística (lint, typecheck, testes) → use pipelines de build/test
- Validação simples de schema → use Zod/JSON Schema direto
- Agent já tem eval harness próprio → use `eval-harness`, `verification-loop`
- Precisa de adversarial review multi-agent → use `santa-method`, `gan-style-harness`

---

## Conceito Central: Dois Níveis de Feedback

| Nível | Quem | Função |
|---|---|---|
| **Execution (baixo)** | Machine/Agent | Mede "quão longe do goal literal" e grinda até zero |
| **Judgment (alto)** | **Human** (ou verifier independente) | Decide "este goal está certo? deve mudar? deve parar?" |

> **Regra de Ouro:** o verifier **nunca** compartilha contexto com o generator. Context window limpo = julgamento independente.

---

## Arquitetura do Padrão

```
GENERATOR → candidate → INDEPENDENT VERIFIER (clean context) → scores + feedback
     ↑                                                                    │
     └────────────── hill-climb até threshold ────────────────────────────┘
                          max_iterations → choose_not_to_show
```

O verifier recebe **só** candidate + rubric + ground truth (se houver). Nunca traces, prompts ou reasoning do generator. Cada critério é avaliado 0–1 com feedback específico; score final = soma ponderada. Ver `references/implementation.md` (`OutcomeRubricVerifier` + `HillClimbLoop` + exemplo de briefs).

---

## Rubric Design Principles (Do vídeo)

| Princípio | Aplicação |

Attribution

majinmagrosmajinmagros
View sourceMore from majinmagros →
SSkills DirectorySkills Directory

Ship a skill? Prove it's safe.

Free 120-pattern security scan, letter grade, and an embeddable README badge.

Submit a skill

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments (0)

No comments yet. Be the first to comment!

SSkills DirectorySkills Directory

Ship a skill? Prove it's safe.

Free 120-pattern security scan, letter grade, and an embeddable README badge.

Submit a skill

Related Skills

Caveman

Ultra-compressed communication mode. Cuts token usage ~75% by speaking like caveman while keeping full technical accuracy. Supports intensity levels: lite, full (default), ultra, wenyan-lite, wenyan-full, wenyan-ultra. Use when user says "caveman mode", "talk like caveman", "use caveman", "less tokens", "be brief", or invokes /caveman. Also auto-triggers when token efficiency is requested.

1023331 votes

Hyperplan

Adversarial multi-agent planning skill. Self-orchestrates 5 hostile category members (unspecified-low, unspecified-high, deep, ultrabrain, artistry) via team-mode for ruthless cross-critique debate, distills only the defensible insights, then MANDATORILY hands the distilled insight bundle to the `plan` agent for executable plan formalization. Use when planning needs maximum rigor and surfacing of weak assumptions, blind spots, and over-engineering. Triggers: 'hyperplan', 'hpp', '/hyperplan', ...

686011 votes

Mcp Code Execution

Routes multi-tool workflows through MCP servers for large datasets and pipelines. Use when Bash tool overhead is limiting throughput on data-heavy tasks.

3331 votes

catchup

Recovers prior coding-agent session context by running `catchup <agent> --since-compact`, which extracts a clean summary of a previous Codex, Claude Code, Antigravity, OpenCode, or Pi Agent session. Use when the user says "catch up", "what did the last session do", "get me up to speed", "I switched agents", or asks to recover/summarize a previous session before continuing. Do NOT use for the current conversation, git history, or any non-agent log.

611 votes

math-skill

A comprehensive mathematical reasoning skill for AI assistants — handles arithmetic to research-level problems with rigorous step-by-step reasoning, systematic verification, and transparent uncertainty handling

381 votes
View all in ai-agents →