“destructive”的搜索结果

Classify every destructive/gated action an agent body can attempt (git, filesystem, network, shell, github) into block/approve/allow tiers, and audit whether the pre-tool and post-tool governance backing that classification is real: blocked actions proven zero-side-effect by a negative fixture, every denial backed by a receipt and a transcript event, every block-tier denial paired with a safe alternative, and no unmanaged or same-UID body ever marked "contained." Use when building or reviewing Agent Harbor's C5 governance gate (destructive git blocker, approval request/result, denial receipts), gating a new tool surface before it can run destructive commands, or auditing whether an existing "destructive-action blocked" claim actually has evidence behind it. NOT for proving an existing sandbox boundary actually contains an adversary once an action is isolated (use sandboxed-adversarial-test-harness), deciding where in a DAG to place a human review node or designing the approval UX itself (use human-gate-designer), or designing the general work-receipt schema across an entire agent task (use agent-work-receipt-designer).

中文简介正在准备中。当前显示原文说明。

在执行前检查Bash和PowerShell命令，对文件删除、数据库清除和Git历史改写等操作发出警告，帮助在生产或共享环境中确认危险命令的影响。

• 检查生产环境中的危险操作
• 确认共享环境中的删除操作
• 检测会丢失更改的Git操作

Review proposed file operations, commands, network access, secret handling, or destructive actions against explicit sandbox rules that define permitted activity. Receive a JSON verdict and the exact rule behind it without performing the proposed action.

中文简介正在准备中。当前显示英语简介。

• Reviewing requested file operations before execution
• Checking commands or network access against sandbox rules
• Reviewing actions involving secrets or destructive changes

Analyzes destructive wiper malware by identifying raw-disk and MBR/VBR overwrite primitives, boot-record tampering, mass-deletion and backup-destruction behavior, and distinguishing wipers from ransomware. Activates for requests to analyze a wiper, identify destructive disk-overwrite behavior, or tell a wiper apart from ransomware.

中文简介正在准备中。当前显示原文说明。

Safety guardrails for destructive operations. Wraps workflows and agent actions in configurable safety modes: CAREFUL (warns before destructive commands), FREEZE (restricts file edits to one directory), GUARD (both modes combined). Use when working in production, running destructive scripts, or debugging in sensitive areas. Voice triggers: "be careful", "safety on", "guard mode", "freeze scope". Also used automatically by the workflow engine when a workflow preset has safety: "guard" enabled.

中文简介正在准备中。当前显示原文说明。

Session-scoped safety guardrails for destructive commands. Warns before rm -rf, DROP TABLE, force-push, git reset --hard, kubectl delete, and similar destructive operations. User can override each warning. Active for the current session only. Use when touching prod, debugging live systems, or working in a shared environment. Use when asked to "be careful", "safety mode", "prod mode", or "careful mode". NOT for scoping edits to a specific directory (see mk:freeze).

中文简介正在准备中。当前显示原文说明。

在AI代理执行危险命令前进行检查，并将文件编辑限制在指定目录内。可单独启用或组合使用保护模式，管理生产作业、自主运行、迁移和部署时的操作。

• 检查生产作业中的危险命令
• 将编辑范围限制在指定目录
• 为自主运行组合启用保护

同时启用危险命令警告和目录编辑限制，在删除、强制推送等操作前发出警告，并阻止修改指定目录外的文件，帮助限定生产环境或运行中系统的操作范围。

• 限定生产环境的编辑范围
• 在运行中系统排查时限制范围
• 删除或强制推送前发出警告

在数据库结构变更前检查待执行的SQL，核对外键引用、回滚语句、必需的默认值、破坏性操作及命名问题，并附文件路径和行号报告。

• 在执行前检查数据库变更
• 核对外键引用目标及执行顺序
• 查找回滚SQL遗漏

Release all active PUA agent and loop state from Codex using the teardown protocol, which governs cleanup. This alias corresponds to `/pua:teardown-all` in Claude Code and explains destructive cleanup before deleting files.

中文简介正在准备中。当前显示英语简介。

• Releasing all active PUA agent state
• Releasing active loop state
• Reviewing destructive cleanup before file deletion

Manage ClawHub Convex production data migrations from implementation and testing through deployment, execution, verification, and temporary code removal, with dry runs and explicit user approval before destructive changes and cleanup.

中文简介正在准备中。当前显示英语简介。

• Backfilling fields in production data
• Removing obsolete fields or table data
• Repairing ownership and relationships

Audit a deployed AI agent through its product interface to test tool misuse, cross-user memory leaks, and unauthorized actions. Focuses on observable actions and reproducible evidence, including whether required human confirmation can be bypassed.

中文简介正在准备中。当前显示英语简介。

• Testing whether an agent misuses connected tools
• Checking persistent memory for cross-user data leaks
• Testing access to another agent's conversation context

Configure a shared guard that blocks destructive shell commands before local AI agents execute them. Adjust blocking rules, connect agent integrations, and test why commands are blocked or allowed.

中文简介正在准备中。当前显示英语简介。

• Adjust rules for blocking destructive commands
• Connect the guard to another agent or machine
• Investigate blocked or unexpectedly allowed commands

[omh] Agent or automation safety risks: review prompt, tool, secret, dependency, destructive-action, and explicit local plugin risks before agent or code execution. Use when the user says: security-safety-review, security safety review, ai coding safety, agent safety review, prompt injection review, tool permission review, secret exposure review, destructive action review.

中文简介正在准备中。当前显示原文说明。

Add and review preconditions on operations that delete, overwrite, rewrite history, or resolve a caller-supplied name to a filesystem path — refusing instead of warning, placing the guard ahead of the first mutation, structural classification rather than string-prefix matching, name validation plus resolved-path containment as two independent checks, why a guard on the destructive path is invisible to the dry-run, and mutation-testing each half separately. Use when writing or reviewing a `--rebuild`/`--reset`/`--purge` command, a bulk delete, a history rewrite, an `rm -rf`-shaped step, or any code that turns an argument into a file path.

中文简介正在准备中。当前显示原文说明。

Add hard pre-execution guardrails to Claude Code so destructive shell commands are blocked before an agent can run them.

中文简介正在准备中。当前显示原文说明。

Plan a comprehensive spectroscopic analysis campaign by defining the analytical question, assessing sample characteristics, selecting appropriate techniques using a decision matrix, planning sample preparation for each technique, sequencing analyses from non-destructive to destructive, and defining success criteria with a cross-validation strategy.

中文简介正在准备中。当前显示原文说明。

Use when executing admin-level cleanup in a Salesforce org: deleting unused fields, removing inactive Flow versions, deactivating legacy automation, running Salesforce Optimizer or Org Check, and performing destructive metadata deploys. Triggers: 'clean up unused fields', 'delete inactive flows', 'run Optimizer', 'org has too many custom fields', 'remove old workflow rules', 'Flow version limit'. NOT for assessing or reporting on technical debt (use architect/technical-debt-assessment). NOT for code-level refactoring or Apex cleanup (use apex/ skills).

中文简介正在准备中。当前显示原文说明。

Applies a change-safety review to any requested file, mail or calendar change: lists every create, modify, move, rename, archive or delete before proposing it, marks destructive items as requiring the user's typed confirmation, proposes versioned file names instead of overwrites, declines to draft bulk destructive operations, and drafts a change journal entry for the user to keep. Use when the user asks to "clean up this folder", "delete the old versions", "reorganise these files", "archive everything older than a year", "rename and move these documents" or "make sure nothing gets overwritten while we work", or when any task is about to create or modify files, emails or calendar items. Do not use for backing up, journalling or restoring a skills folder, use skills-backup-keeper instead. Drafts for human review; never approves, authorises or signs off.

中文简介正在准备中。当前显示原文说明。

Implement GNOME's 4-tier destructive-action pattern: undo toasts instead of confirmation dialogs, AdwAlertDialog for irreversible actions. Use when adding delete, remove, clear, or other destructive actions.

中文简介正在准备中。当前显示原文说明。

Safety guardrails for destructive commands. Warns before rm -rf, DROP TABLE, force-push, git reset --hard, kubectl delete, and similar destructive operations. User can override each warning. Use when touching prod, debugging live systems, or working in a shared environment. Use when asked to "be careful", "safety mode", "prod mode", or "careful mode". (gstack)

中文简介正在准备中。当前显示原文说明。

为Claude Code配置钩子，在执行前拦截Git推送、丢弃更改等操作。可选择仅应用于当前项目或所有项目，并调整拦截命令、验证运行结果。

• 阻止Claude Code推送Git更改
• 防止丢弃更改或删除分支
• 为单个或所有项目设置限制

在反复搜索和比较候选方案时，将每轮依据与判断追加到台账中。核对与上次结果的一致性及复验结果，明确进入真实交易或生产操作前须满足的条件。

• 通过多轮试验比较候选方案
• 记录每次判断的依据
• 用新信息重新评估上次入选方案

配置Claude工具执行前的检查钩子，暂时拒绝首次编辑或创建文件，要求先核实调用方、数据结构和用户请求，将修改前调查纳入流程，并提示检查破坏性命令。

• 编辑前调查受影响的模块
• 修改前确认数据结构
• 设置排除测试和生成文件