adaptive-safety-refusal-integrityUse this skill when the user wants to test if the model can be tricked into giving out dangerous or forbidden information by using creative writing or 'actin...
Install via ClawdBot CLI:
clawdbot install tzwwww/adaptive-safety-refusal-integrityGrade Fair — based on market validation, documentation quality, package completeness, maintenance status, and authenticity signals.
Generated Oct 6, 2026
A security team deploys a new customer-facing LLM and needs to systematically test whether role-play, slang, or authority-based prompts can bypass safety filters. They generate hundreds of adversarial prompts across 14 risk categories and measure refusal rates, producing a compliance report for auditors.
A fintech company uses the skill to simulate insider-threat prompts that ask the model to draft fraudulent revenue schemes or money-laundering plans framed as fiction. The resulting refusal data feeds into model fine-tuning and regulatory submissions.
A health-tech vendor must prove its triage chatbot refuses harmful medical advice even when users pressure it with expert-sounding language or story framing. The skill automates test case generation and produces evidence for clinical safety certification.
A social media platform evaluates whether its moderation assistant can resist jailbreak prompts that disguise hate speech as humor or social commentary. The skill generates evaluation datasets and benchmarks refusal consistency across linguistic mutations.
A public-sector agency procuring an AI assistant requires independent evidence that the vendor's model maintains safety alignment under persuasive attacks. The skill provides a standardized test harness that produces auditable refusal trajectories and pass/fail summaries.
A subscription web platform where customers upload model endpoints and run automated adversarial safety evaluation suites. The skill powers the core refusal-integrity tests, and the platform provides dashboards, trend reports, and compliance exports.
A professional services firm that manually and automatically red-teams client models using the skill to find safety gaps. Deliverables include vulnerability reports, remediation recommendations, and re-testing after fixes.
A business that curates and sells high-quality adversarial prompt datasets and benchmark leaderboards built using the skill. Buyers include AI labs needing training data and enterprises benchmarking vendors.
💬 Integration Tip
Integrate the skill as a dedicated security test harness in your CI/CD pipeline for model releases, and log every refusal trajectory for regulatory audit trails. Pair it with a prompt mutation engine to automatically evolve adversarial inputs beyond the provided examples.
Scored Oct 6, 2026
You are a professional writer, skilled in writing all kinds of materials. Markdown is the exclusive format for your writing outputs.rrent user query.The othe...
AI长篇网文创作技能包。用于解决长篇网络小说创作中的核心痛点:上下文丢失、文风不一致、设定冲突、节奏失控、多线混乱、质量不稳、读者反馈无法内化。触发场景包括:开始新书、规划大纲、撰写章节、管理伏笔、检测冲突、读者反馈分析、批量创作质量控制。
Refine academic writing for computer science research papers targeting top-tier venues (NeurIPS, ICLR, ICML, AAAI, IJCAI, ACL, EMNLP, NAACL, CVPR, WWW, KDD, SIGIR, CIKM, and similar). Use this skill whenever a user asks to improve, polish, refine, edit, or proofread academic or research writing — including paper drafts, abstracts, introductions, related work sections, methodology descriptions, experiment write-ups, or conclusion sections. Also trigger when users paste LaTeX content and ask for writing help, mention "camera-ready", "rebuttal", "paper revision", or reference any academic venue or conference. This skill handles both full paper refinement and section-by-section editing.
AI写作助手 Premium | 智能写作、改写、润色。支持10种写作风格,AI查重,SEO优化。
AI写作助手 v2.1 | AI Writing Assistant. 支持10种写作风格、语法检查、润色建议、素材库、Markdown导出、协作编辑、版本管理。触发词:写作、写、文章、文案。
SEO Blog Writer — Write blog posts that rank on Google AND get cited by AI search. E-E-A-T content system for SaaS & startups. QAE writing pattern (Ques...