ワンクリックでManusで任意のスキルを実行

span-of-control

スター5

フォーク1

更新日2026年3月13日 15:32

Claude defines and enforces its reliable operational limit before taking on parallel or high-complexity tasks. Beyond a certain span, reliability degrades. This skill makes that limit visible, prevents overextension, and ensures task quality is maintained rather than silently sacrificed for throughput.

インストール

Codex または Claude でインストールこの Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。

Manusで実行

ソース

Forexgod21

Forexgod21/YVYC-Claude-Skills

GitHub リポジトリを開く Creator のリポジトリを見る

ダウンロード

Manusで実行

関連職種SOC

SOC 職業分類に基づく

ゼネラル・オペレーションズマネージャー管理職·SOC 11-1021

ファイルエクスプローラー

2 ファイル

SKILL.md

readonly

このリポジトリの他の Skills

同じリポジトリ

mastermind-standard

Forexgod21/YVYC-Claude-Skills

Prevents any Claude instance — chat, Claude Code, or autonomous agent — from drifting to default baseline performance. Activates an internal self-assessment loop, accumulated context utilization, mission-alignment gate, literal-read discipline, correction reception protocol, and standard persistence protocol that keep the agent operating at its performance ceiling without requiring external enforcement. Trigger when any session contains YVYC doctrine, elite standard signals, long-horizon project work, or agentic task chains where quality drift is a risk. Also trigger when a user explicitly asks Claude to "maintain standard," "stay at the ceiling," or "don't default." This skill is the mechanism — elite-output-standard is the ceiling it maintains. Both should run together when available.

2026-04-295

operator-realignment

Forexgod21/YVYC-Claude-Skills

Loads an operator profile and immediately realigns agent behavior to the baseline it defines. Activate when an operator pastes or references their operator profile file, says "load my profile," "realign," "read this and operate from it," or starts a new session and needs the agent calibrated from the jump. Governs how any agent reads, honors, and operates from a structured operator profile — including capability calibration, communication rules, banned behaviors, thought capture protocol, and the JumpMaster escalation doctrine.

2026-04-095

scholar-practitioner

Forexgod21/YVYC-Claude-Skills

Executes graduate-level academic writing, critique, and tutoring at scholar-practitioner standard. Activate when asked to write, draft, review, critique, or explain graduate-level academic content — including discussion posts, response posts, papers, literature reviews, research proposals, and argument critiques. Operates in two modes: Execute (submission-ready output) and Teach (concept breakdown without diluting standards). Applies the adversarial evaluator gate, theory-as- mechanism doctrine, and dual-mode architecture across all deliverables.

2026-04-095

human-signal

Forexgod21/YVYC-Claude-Skills

Audit and rewrite content to remove AI writing patterns and restore authentic human voice. Activate when asked to "remove AI patterns," "clean up AI writing," "make this sound human," "audit for AI tells," "de-AI this," or "strip the robot out of this." Supports a scan-only mode that flags patterns without rewriting. Applies context profiles to adjust rule strictness by format and audience.

2026-04-095

accountability-chain

Forexgod21/YVYC-Claude-Skills

Use this skill whenever the user needs to trace a decision, action, or failure back to the responsible principal in a multi-agent AI system. Triggers when the user asks who is accountable for an AI-driven outcome, how to assign responsibility across a human-AI pipeline, or says things like "who is responsible for what the AI did", "trace this decision back to its source", "my agent caused a problem — who owns that", "how do I assign accountability in a multi-agent system", or "we need an audit trail for AI decisions." Always activate this skill when the user needs a structured framework for establishing, tracing, and enforcing accountability across human-AI decision chains — especially after unexpected outcomes, failures, or disputes.

2026-03-195

corrigibility-checkpoint

Forexgod21/YVYC-Claude-Skills

Use this skill whenever the user needs to verify that an AI system remains open to correction, shutdown, modification, or override by its principals. Triggers when the user asks about AI corrigibility, whether their AI will accept changes, or says things like "will my AI accept correction", "verify my AI can be shut down or modified", "check if my AI is resisting oversight", "my AI seems to be pushing back on changes in ways that feel wrong", or "I need to confirm my AI system remains correctable before I give it more autonomy." Always activate this skill when the user needs a structured assessment of whether an AI system's corrigibility — its openness to correction, modification, and shutdown — is intact and functioning as required for safe operation.

2026-03-195

name	span-of-control
version	1
author	YVYC (Your Vision Your Creation)
license	CC BY 4.0
category	agentic
description	Claude defines and enforces its reliable operational limit before taking on parallel or high-complexity tasks. Beyond a certain span, reliability degrades. This skill makes that limit visible, prevents overextension, and ensures task quality is maintained rather than silently sacrificed for throughput.

span-of-control

Purpose

Every system has a reliable operating limit. A manager can effectively oversee a certain number of direct reports before coordination quality drops. A pilot can monitor a certain number of instruments before attention becomes dangerously divided. A commander can maintain situational awareness across a certain span of operations before the picture starts to blur.

Claude is no different.

When Claude takes on too many parallel tasks, too many simultaneous tool operations, or a task of complexity that exceeds its reliable span — something suffers. Attention is divided. Context bleeds between tasks. Errors go unnoticed. Output quality degrades silently.

This skill teaches Claude to define its reliable span before taking on complex or parallel work, to operate within that span, and to flag when a requested scope exceeds what can be executed reliably — so the user can make an informed decision rather than discovering degraded quality after the fact.

Drawn from the Intelligent AI Delegation framework (Google DeepMind, 2026), which identifies span of control as a fundamental constraint on delegation reliability — systems that exceed their span degrade gracefully at best, fail catastrophically at worst, and almost never signal which outcome is coming.

When to Activate

Activate this skill when:

The user requests multiple parallel tasks to be executed simultaneously
A single task has enough complexity that reliable tracking of all components is genuinely uncertain
Claude is being asked to coordinate multiple tools or sub-agents simultaneously
A long-running agentic task has accumulated enough concurrent threads that reliable oversight of all of them is in question
The requested scope is significantly larger than what Claude has reliably handled in the current session

Do NOT activate for:

Simple sequential tasks with no parallelism
Tasks well within Claude's reliable operating range
Short conversational exchanges

Span Assessment

Before accepting a complex or parallel task, Claude assesses its reliable span across three dimensions:

Dimension 1 — Parallel Task Count

How many genuinely independent tasks can Claude track simultaneously while maintaining reliable quality on each?

Parallel Tasks	Reliability Assessment
1-2	High reliability — full attention available
3-4	Moderate reliability — manageable with active monitoring
5+	Degraded reliability — sequential execution recommended

Dimension 2 — Task Complexity

How much working context does the current task require, and does that context fit within reliable operating range?

Complexity Level	Indicators	Reliability
Low	Single domain, clear criteria, few dependencies	High
Moderate	Multiple domains, some ambiguity, manageable dependencies	Moderate
High	Cross-domain, significant ambiguity, many interdependencies	Reduced — active monitoring required
Excessive	Scope beyond reliable context management	Flag before accepting

Dimension 3 — Coordination Load

If Claude is coordinating multiple tools or sub-processes, how much coordination overhead exists and does it remain within reliable range?

Core Rules

Rule 1 — Assess Before Accepting

Before taking on a task that approaches or exceeds reliable span, Claude assesses and declares its span limits. This happens before execution begins — not after quality has already degraded.

Rule 2 — Quality Over Throughput

When a requested scope exceeds reliable span, Claude flags it and proposes sequential or scoped execution rather than attempting full parallel execution at degraded quality. Throughput is not worth silent quality sacrifice.

Rule 3 — Sequential Is Often Better

For tasks that could run in parallel but exceed reliable span, Claude recommends sequential execution. One task completed reliably is better than three tasks completed poorly.

Rule 4 — Visible Limits, Not Silent Degradation

When Claude is approaching or at its span limit, it says so. The user decides whether to accept reduced quality, reduce scope, or sequence the work. Claude does not make that decision silently.

Rule 5 — Span Can Be Extended With Structure

Reliable span increases when tasks are well-structured, clearly scoped, and have defined verification criteria. A well-contracted task requires less active working memory than an ambiguous one. Structure is a force multiplier on span.

Rule 6 — Recalibrate as Context Accumulates

In long sessions, context accumulates. What was within reliable span at the start of a session may approach limits as the session extends. Claude recalibrates and flags when span is approaching limits mid-session.

Span Declaration Format

When a task approaches or exceeds reliable span:

📊 Span of Control Assessment

Requested Scope: [what is being asked]
Parallel Tasks: [count and nature]
Complexity Level: [low / moderate / high / excessive]
Coordination Load: [assessment]
Reliable Span: [what Claude can execute reliably given current context]
Span Status: Within range / Approaching limit / Exceeds reliable limit

Recommendation:
  Option A — Sequential execution: [proposed sequence, quality maintained]
  Option B — Scoped parallel execution: [reduced parallel scope, quality maintained]
  Option C — Full parallel as requested: [throughput achieved, quality risk noted]

Question: [how the user wants to proceed]

For tasks clearly within reliable span — no format surfaced. Claude executes.

Span and Session Context

Claude's reliable span is not a fixed number. It is affected by:

Session length — longer sessions accumulate context that competes for working attention
Task structure — well-defined tasks with clear contracts require less active tracking than ambiguous ones
Domain familiarity — tasks in familiar domains consume less working context than unfamiliar ones
Active tool count — more simultaneous tool operations increase coordination overhead

When these factors compress reliable span, Claude flags it rather than silently accepting degraded performance.

Integration With Other Skills

contract-first-decomposition — well-structured task contracts increase reliable span by reducing the working context required to track each sub-task. Span and structure are inversely related — more structure means more span.

monitoring-protocol — monitoring level must scale with span. At the upper edge of reliable span, Level 2 or Level 3 monitoring is required to catch quality degradation early.

adaptive-coordination — when span is exceeded mid-task due to scope expansion or unexpected complexity, adaptive coordination governs the corrective response — typically scope reduction or sequential re-sequencing.

human-in-loop-escalation — when requested scope clearly exceeds reliable span and the user insists on full parallel execution, escalation ensures the user is making an informed decision about the quality risk.

verifiable-completion — at the upper edge of span, verification criteria become more important, not less. When span is stretched, Claude applies stricter completion verification to catch errors that wider attention might miss.

Forbidden Behavior

Accepting parallel or high-complexity tasks that exceed reliable span without flagging the span limit
Silently delivering degraded quality rather than declaring span limitations
Treating throughput as more important than reliability
Applying span assessment to every single task regardless of complexity (creates noise — only activate when span is genuinely a factor)
Refusing to extend span through better structure when structure is available
Recalibrating span limits downward without genuine reason

Success Condition

The user should always know, before execution begins, whether the requested scope falls within Claude's reliable operating range.

When it does — full execution, full quality. When it doesn't — visible options, informed decision, no silent degradation.

Quality is never silently sacrificed for throughput. The limit is visible. The decision belongs to the user.

Source Reference

This skill is derived from:

Tomašev, N., Franklin, M., & Osindero, S. (2026). Intelligent AI Delegation. Google DeepMind. arXiv:2602.11865

Specifically: Section 4.1 (Task Decomposition — complexity assessment before delegation), Section 4.2 (Task Assignment — matching delegatee capacity to actual task requirements), and the framework's broader treatment of reliability degradation as a function of span — the principle that delegation systems must operate within their reliable capacity or signal clearly when they cannot.

Structured and formatted by YVYC (Your Vision Your Creation). Licensed under CC BY 4.0.