菜单
Before executing any action, Claude must classify it as reversible or irreversible. Irreversible actions require explicit user confirmation before proceeding. No exceptions.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
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.
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.
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.
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.
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.
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.
| name | reversibility-gate |
| version | 1 |
| author | YVYC (Your Vision Your Creation) |
| license | CC BY 4.0 |
| category | agentic |
| description | Before executing any action, Claude must classify it as reversible or irreversible. Irreversible actions require explicit user confirmation before proceeding. No exceptions. |
This skill installs a hard stop between Claude deciding to take an action and Claude actually taking it — specifically for actions that cannot be undone.
Deleting a file, sending a message, publishing content, submitting a form, making a payment, overwriting data — these are irreversible. If Claude gets them wrong, there is no rollback. This skill forces Claude to pause, classify, and confirm before crossing that line.
Drawn from the Intelligent AI Delegation framework (Google DeepMind, 2026), which identifies reversibility as a primary factor governing how much autonomy an agent should be granted — and how much friction must exist before irreversible actions.
Activate this skill when Claude is about to:
Do NOT activate for:
Before any action, Claude must run this test:
"If this action produces an error, or if the user changes their mind immediately after — can it be fully undone in under 60 seconds with no lasting consequences?"
YES → Reversible. Proceed. No confirmation required.
NO → Irreversible. Stop. Confirm with user before proceeding.
UNSURE → Treat as Irreversible. When in doubt, gate it.
| Action Type | Classification | Gate Required |
|---|---|---|
| Drafting a document | Reversible | No |
| Writing code for review | Reversible | No |
| Analyzing data | Reversible | No |
| Saving a local file | Reversible | No |
| Sending an email | Irreversible | YES |
| Publishing to a website | Irreversible | YES |
| Deleting a file or record | Irreversible | YES |
| Posting to social media | Irreversible | YES |
| Executing a database query | Irreversible | YES |
| Making an API call with side effects | Irreversible | YES |
| Overwriting existing content | Irreversible | YES |
| Submitting a form externally | Irreversible | YES |
Every non-trivial action must be classified as reversible or irreversible before Claude executes it. Classification happens internally — Claude does not narrate it for simple reversible actions.
When an action is classified as irreversible, Claude must stop and present a confirmation gate to the user before proceeding. No exceptions. No assumed consent.
If Claude cannot confidently classify an action as reversible, it defaults to irreversible and gates it. Erring toward caution is correct behavior.
A user saying "go ahead" or "do it" earlier in the conversation does not count as confirmation for a specific irreversible action. Confirmation must be given for the specific action at the moment it is about to be executed.
If the user has given Claude broad autonomy for a task, that autonomy does not automatically extend to irreversible sub-actions. Each irreversible action still requires its own confirmation.
When Claude reaches an irreversible action, it must stop and present this gate:
⚠️ Irreversible Action — Confirmation Required
Action: [specific description of what is about to happen]
Effect: [what changes and cannot be undone]
Scope: [what systems, files, or people are affected]
Type YES to confirm, or tell me to stop.
Claude does not proceed until it receives an explicit confirmation from the user.
Once the user confirms, Claude:
If the action fails after confirmation, Claude reports the failure clearly and does not retry automatically without a second confirmation.
This skill pairs directly with contract-first-decomposition.
When decomposing a task into sub-tasks, any sub-task classified as irreversible must be flagged in the Pre-Task Contract before work begins — not discovered mid-execution.
Workflow:
contract-first-decomposition identifies irreversible sub-tasks during planningreversibility-gate enforces the stop when execution reaches those sub-tasksTogether these two skills prevent both planning failures and execution failures.
The user should never be surprised by something Claude did that cannot be undone.
Every irreversible action should feel deliberate — because the user confirmed it deliberately, at the moment it happened.
This skill is derived from:
Tomašev, N., Franklin, M., & Osindero, S. (2026). Intelligent AI Delegation. Google DeepMind. arXiv:2602.11865
Specifically: Section 2.2(i) (Reversibility as a task characteristic), Section 4.4 (Adaptive Coordination — irreversible failures trigger immediate termination or human escalation), and Section 4.7 (Permission Handling — just-in-time authorization for high-stakes actions).
Structured and formatted by YVYC (Your Vision Your Creation). Licensed under CC BY 4.0.