“guardrails and AI safety”

配置跨平台、宿主中立的本地 Git 防破坏拦截（阻断强推、硬重置、强制清理与强删分支等不可逆操作）。

A constraint you write into a brief for a subagent or a future self binds you first; authoring a frame can make it invisible to you.

Paid media budget, kill, and scale rules tied to contribution margin.

Installs guardrails in this project's .claude/ (and its git worktrees) that keep Claude Code from reading, writing or running commands on paths outside a…

Soft rules for Claude to avoid dangerous git commands (force-push, hard reset, branch -D).

Use this skill when you need evidence-bounded safety policy, abuse categories, refusal/redirect behavior, privacy, escalation, and Human risk decisions…

Use when page, component, form, table, dashboard, responsive layout, or UI review work needs product-quality implementation guardrails.

AI-guardrail linting — size/complexity caps that force extraction over sprawl, no comments, named constants, everything an error, and the config protected…

Configure and test agent guardrails for input/output validation, tool-use constraints, and safety boundaries

Set up a Copilot hook to block dangerous git commands (push, reset --hard, clean, branch -D, etc.) before they execute.

Verify git remote safety before push: correct remote URL, correct branch, no force push without verification.

Set up git safety guardrails to block dangerous git commands (push, reset --hard, clean, branch -D, etc.) before they execute.

Use when making Git changes while preserving user work and repository history.

One shared denylist of catastrophic shell commands (rm -rf on / or ~, dd/mkfs, sudo rm, fork bombs, curl|sh, git push --force, gh repo delete) enforced as a…

Avoid the common analysis traps when answering revenue questions — grain, dates, and default filters.

Enforce command approval, secret redaction, and PII hashing rules.

Audits an agent workflow or config for missing guardrails - sandbox scope, approvals, a give-up limit, a tool allowlist, read-only review, and network egress.

Set up Claude Code hooks to block dangerous git commands (push, reset --hard, clean, branch -D, etc.) before they execute.

Añade guardrails de Git para sesiones de Codex mediante instrucciones y un verificador local.

ELYTH/からくり発話やmemory proposalのsecret・cross-surface混入を検査する。

Disposable meta-skill (delete after the harness is built): sets up a GitHub repository's platform-enforced guardrails — automated dependency updates via…

Use when installing, auditing, or testing executable safety hooks for GuanFu, especially to block dangerous git commands such as push, reset --hard, clean -f…

Implement comprehensive safety guardrails for LLM applications including content moderation (OpenAI Moderation API), jailbreak prevention, prompt injection…

Respond to destabilization, crisis signals, or over-intensified practice by simplifying and redirecting to appropriate support.