Agent Skills

---

name: resemble-detect

description: Deepfake detection and media safety — detect AI-generated audio, images, video, and text, trace synthesis sources, apply watermarks, verify speaker identity, and analyze media intelligence using Resemble AI

---

name: skill-creator

description: Guide for creating effective skills. This skill should be used when users want to create a new skill (or update an existing skill) that extends Claude's capabilities with specialized knowledge, workflows, or tool integrations.

---

name: slack-gif-creator

description: Toolkit for creating animated GIFs optimized for Slack, with validators for size constraints and composable animation primitives. This skill applies when users request animated GIFs or emoji animations for Slack from descriptions like "make me a GIF for Slack of X doing Y".

---

name: template-skill

description: Replace with description of the skill and when Claude should use it.

---

name: theme-factory

description: Toolkit for styling artifacts with a theme. These artifacts can be slides, docs, reportings, HTML landing pages, etc. There are 10 pre-set themes with colors/fonts that you can apply to any artifact that has been creating, or can generate a new theme on-the-fly.

---

name: video-downloader

description: Downloads videos from YouTube and other platforms for offline viewing, editing, or archival. Handles various formats and quality options.

---

name: webapp-testing

description: Toolkit for interacting with and testing local web applications using Playwright. Supports verifying frontend functionality, debugging UI behavior, capturing browser screenshots, and viewing browser logs.

---

name: computer-use

description: Use when a task needs an app on the user's own screen — a native desktop application (instrument control software, Origin, GraphPad, ImageJ, a reference manager, a spreadsheet), or a desktop browser window that needs window-level control. Read before the first `computer` call. Covers the observe-act-observe loop, why element indexes go stale, which action to prefer for which control, what "unverified" means and why it must never be reported as success, and the recovery for each error code. Not for page-level web automation (use the browser connector) and not for anything that can be done from the shell.

---

name: domain-check

description: Use ONLY when the user asks to check, validate, review or audit analysis code for scientific correctness, or when you are the reviewer agent running its checks. Never run it unprompted as part of ordinary analysis work. Covers physics, earth/geo, biology, chemistry, social science and bioprocess/fermentation. Runs a deterministic domain-correctness gate that catches code which runs but is scientifically wrong (unit/dimension mismatch, Euclidean distance on lat/lon without a CRS, 0-based/1-based coordinate and strand errors, impossible SMILES valence, uncorrected multiple comparisons, averaging a categorical code, unconstrained kinetic-parameter fits, kLa computed from raw DO instead of the driving force, arithmetic mean of raw CFU counts, ANOVA/Tukey with no assumption check, a curvature DOE design fit with a first-order model, an ML model fit across process scales with no correction). Surfaces structured findings; never claims the code is correct.

---

name: large-file

description: Use BEFORE reading any data file that could be large (CSV/TSV, Parquet, HDF5, FITS, NetCDF, NDJSON, genomics FASTQ/FASTA/VCF/BAM, GRIB, ROOT, or big text/simulation logs like VASP OUTCAR). Returns a compact memory pointer — header/schema/shape/sample/key numbers — by introspection and sampling in bounded memory, so you never load a file bigger than the context window into the model. Reference data via the pointer; read specific ranges deterministically.

---

name: modal-run

description: Use when the user asks to run heavy or GPU work on Modal (the cloud compute platform) — writing a Modal function in the workspace, running it with the user's own `modal` CLI + token, and bringing results back. Data-to-compute for jobs too big for the laptop, without a Slurm cluster.

---

name: publication-figures

description: Use whenever you generate or review a chart, plot, table, or paper figure in this workspace, including work delegated by paper-writing, literature-survey, and experiment skills. Applies the Open Science publication style, enforces readable final-size layout for figures and tables, and rejects generic diagram-tool output as a publication figure. Interactive Plotly/HTML may be used for exploration, but paper delivery requires a static publication-ready export.

---

name: remote-compute

description: Use when the user asks to run, submit, monitor, or cancel a job on a remote machine over SSH — their own GPU/CPU server, a workstation, or a Slurm cluster ("the cluster", a login node, "my 3090 box", "the compute server"). Picks a saved machine, runs the work directly over SSH (or via Slurm when present), tracks it, and fetches results back into the workspace.

---

name: stats-integrity

description: Use whenever you run statistical analysis for the social sciences (regression, hypothesis tests, econometrics) or read Stata (.dta) / SPSS (.sav) data in this workspace. Enforces an execute-don't-interpret boundary (surface estimates, don't volunteer causal claims), checks the analysis against a preregistration plan for HARKing, verifies reproducible seeds, and reproduces .dta/.sav estimates via R. Flags integrity risks; never certifies the analysis is sound.

---

name: traceability-review

description: Use when the user asks to review, verify, or audit a report, manuscript, or analysis in the workspace for traceability — resolving citations, flagging numbers with no source, and checking figures against the code that generated them. Emits a structured review block the app renders as reviewer findings. Verifies traceability, never "correctness".

---

name: claude-reflect

description: Self-learning system that captures corrections during sessions and reminds users to run /reflect to update CLAUDE.md. Use when discussing learnings, corrections, or when the user mentions remembering something for future sessions.

---

name: agent-security

description: AI Agent 系统安全测试方法论。当目标系统使用 AI Agent 执行工具调用、多 Agent 协作、 或自主决策时触发。覆盖 OWASP Agentic AI Security Top 10 (ASI01-ASI10): 目标劫持、工具滥用、身份权限、供应链、代码执行、记忆投毒、多 Agent 通信、 级联故障、人机信任利用、失控 Agent。

---

name: ai-data-security

description: AI 系统数据安全测试方法论。当需要评估 LLM/AI 系统的数据泄露风险、训练数据安全、 或 RAG/向量库数据完整性时触发。覆盖: System Prompt 泄露(元 Prompt/角色扮演/关键字定位)、 训练数据推导与提取、成员推断攻击、模型反演攻击、RAG 数据投毒、API 信息泄露、 级联幻觉攻击、外部数据源信息泄露。

---

name: ai-identity-security

description: AI 系统身份与权限安全测试方法论。当目标系统涉及 Agent 身份认证、多 Agent 权限管理、 角色设定安全、会话管理、或 MCP/API 凭据管控时触发。 覆盖: 角色逃逸(假定场景/假定角色/遗忘法/目标劫持)、权限失控(Action 越权/MCP 未授权资源获取)、 多 Agent 身份伪造、会话劫持、凭据泄露与滥用。

---

name: aliyun-pentesting

description: 阿里云渗透测试方法论。当目标使用阿里云服务、发现 oss-*.aliyuncs.com 资产、获取阿里云 AccessKey（LTAI 开头）、在 ECS 实例内部可访问 100.100.100.200 元数据、或需要对阿里云 RAM/ECS/OSS/RDS/ACK/FC 等服务进行安全评估时使用。覆盖 RAM 提权、ECS 接管、OSS Bucket 利用、RDS 数据库攻击、ACK 容器集群、FC 函数计算、SLB 负载均衡、日志服务、KMS 密钥管理

---

name: aws-iam-policy-analysis

description: AWS IAM / Resource Policy 分析方法论。当目标涉及 AWS 云服务且提供了 IAM Policy、Resource Policy、Lambda 代码、CloudFormation 模板等策略文件时使用。覆盖 IAM Policy 危险模式识别、Resource Policy Principal 分析、Condition 键绕过模式、服务信任关系图推导。发现 AWS 策略文件（json/yaml）、*.amazonaws.com 端点、IAM ARN 时应使用此技能

---

name: aws-iam-privesc

description: AWS IAM 权限提升专项方法论。当已获取 AWS 凭据并需要提升权限、发现当前 IAM 用户/角色权限有限需要横向或纵向提权、或需要分析 IAM Policy 寻找提权路径时使用。覆盖 46 个 AWS 服务的提权技术，包括 PassRole 滥用、AssumeRole 链式提权、Lambda/EC2/ECS 计算服务提权、以及 NotAction 隐式权限利用

---

name: aws-pentesting

description: AWS 云环境渗透测试总体方法论。当目标使用 AWS 云服务、发现 AWS 相关资产（S3 Bucket/EC2 实例/Lambda 函数/CloudFront 分发）、获取 AWS 凭据（AK/SK/Session Token/IAM Role）、或需要对 AWS 环境进行安全评估时使用。提供从未授权枚举到提权、后渗透、持久化的全流程攻击决策树，引导到专项技能深入。覆盖 47+ AWS 服务攻击面

---

name: aws-post-exploit

description: AWS 后渗透与持久化专项。当已获取 AWS 高权限（Admin/PowerUser）或特定服务权限后，需要进行数据窃取、横向移动到其他账户/区域、建立持久化后门、或清理攻击痕迹时使用。覆盖 S3/RDS/DynamoDB/EBS 数据窃取、跨账户 AssumeRole 横向、Lambda/IAM/EventBridge 持久化后门、以及 CloudTrail/GuardDuty 规避技术