设计与多媒体

human-gates

Part of the Overpowered skill suite. Design or review where human approval, judgment, or accountability must remain in an automated workflow. Use when automation crosses authority, legal, financial, safety, irreversible, or genuinely ambiguous decisions; do not preserve manual steps merely because they exist today.

它能做什么

技能文档

Human Gates

Core rule

Keep humans for authority, accountability, material ambiguity, and unacceptable risk — not for work a deterministic or agentic system can reliably perform.

Procedure

For each candidate human step:

Identify why the human is currently involved.
Classify the reason:
- authority — only an authorized person may decide;
- accountability — an accountable owner must accept the outcome;
- material ambiguity — available evidence cannot reliably determine the case;
- risk control — error cost exceeds the accepted automation boundary;
- legacy/manual habit — no durable reason remains.
Automate legacy/manual steps when controls are sufficient.
Define each retained gate with:
- trigger condition;
- decision owner or role;
- evidence presented to the human;
- allowed decisions;
- timeout/delegation path;
- downstream effect.

Output contract

Gate | Trigger | Why human | Required evidence | Owner | Outcomes

Gotchas

“Human in the loop” is not a design; define the exact gate.
Do not make a human re-check deterministic calculations unless policy requires it.
Historical manual ownership is not proof that human judgment is needed.
Never infer delegated authority from precedent alone.

Stop condition

Stop when every retained human step has a concrete reason and every removed human step has an adequate automated control.

human-in-loop-review

给自主智能体/自动化流水线加一层「人类在环审核」：把中高危、不可逆、外发隐私、提权、支付等有副作用的动作路由到人工审核队列，阻止无人值守 agent 越权执行。内置审核分级（needs_review）、审核队列（add/pending/approve/reject/summary）、完整审计（提议者/风险/审核人/结论/理由/时间戳）。与 safety-guardrails 互补：护栏做 ALLOW/CONFIRM/DENY 决策，本技能把 CONFIRM 类动作转人工队列。触发词：人类在环、人工审核、审核队列、审批流、human-in-the-loop、操作确认、越权拦截、agent 审核。

automate-this

Part of the Overpowered skill suite. Transform a described human/business process into an executable automation design by separating deterministic steps, agentic reasoning, existing systems, and human gates; define triggers, data, decisions, exceptions, controls, and verification. Use when the goal is to automate a process, not merely understand it or interview the user about it.

using-overpowered

Part of the Overpowered skill suite. Route knowledge-work and enterprise-automation tasks through the smallest relevant combination of Overpowered skills, and invoke gear-up only when a material execution capability is genuinely missing. Use when multiple suite skills may apply or when starting a non-trivial task involving organizational knowledge, conflicting evidence, evolving artifacts, business rules, process automation, risky side effects, verifiable completion, or an uncovered capability gap.

Ask First, Log Everything - Contro1 Approvals

How to ask a human before a sensitive action and how to log every autonomous action, so your work is accountable. Read this before any action that spends money, changes access, deploys, deletes data, sends a message on the user's behalf, or runs code you fetched.

completion-audit

Part of the Overpowered skill suite. Independently determine whether a claimed task outcome is actually complete by reconstructing completion criteria and checking fresh evidence. Use before declaring work done, accepted, migrated, reconciled, fixed, or delivered; not as a general planning loop.

skillify

Part of the Overpowered skill suite. Generalize and package a qualified or otherwise proven repeatable workflow into a portable Agent Skill with concise instructions, progressive disclosure, trigger-focused metadata, gotchas, and evals. Use after real execution evidence exists—especially for a Skill Academy capability ready to graduate; do not use as the hot-generation mechanism during a task or merely to save a conversation.