LLM Mart Basic

Answers natural-language questions about the user's database. Loads the agami semantic model (subject areas, tables, columns, relationships with join cardinality, entities, metrics) and few-shot examples from <artifacts_dir>/<profile>/, generates SQL via the examples-first traver

Reconciles known (label, expected_value) numbers from an existing dashboard against agami's answers. Input can be a SCREENSHOT of a Metabase / Power BI / Tableau / Looker dashboard (Claude's vision extracts the pairs), a CSV, or numbers pasted inline — the user doesn't need to kn

Saves a user correction so future queries learn from it. Always appends a (question, corrected_sql) pair to the subject area's example library under <artifacts_dir>/<profile>/prompt_examples/<area>/. Additionally, classifies the correction and — when applicable — applies a surgic

Wires the local agami MCP server (python -m mcp_harness) into the Claude Desktop app in one step, so you can ask your database questions from Claude Desktop — not just inside Claude Code. Auto-detects the right Python interpreter (the one with your DB driver), installs the agami-

根据用户描述生成高质量绘图 prompt，并按通用、roadmap、schematic 模式通过 BenszAPI 直接完成 gpt-image-2 或 Nano Banana/Gemini 出图、编辑和多轮迭代；这是自包含的图片生成工作流，选中后不得调用或依赖 imagegen，除非用户明确要求同时使用 imagegen。

当用户明确要求"测试代码"、"运行代码审查"或"进行代码自检"时使用。通过多轮 A 轮批判性代码审查 + B 轮代码质量原则检查，系统化发现、记录、修复程序代码中的问题，并将计划/过程/结果统一沉淀到目标代码根目录的 `.bensz-api/task-{yyyymmdd-hhmm}-{简短描述}/auto-test-code/{yyyy-mm-dd-hh-mm}/output/tests/` 隔离工作区。⚠️ 不适用：用户只是想优化功能（应直接修改）、只是询问代码问题（应直接回答）、没有明确"测试代码"意图。

当用户明确要求"测试项目"、"运行 auto-test-project"或"进行项目级测试"时使用。对完整项目进行多轮 A 轮批判性测试 + B 轮质量检查，系统化发现、记录、修复问题。⚠️ 不适用：用户只是想优化功能（应直接修改）、只是询问项目问题（应直接回答）、没有明确"测试"意图。

当用户明确要求"测试技能"、"运行 auto-test"或"进行批判性测试"时使用。通过多轮 A 轮批判性测试 + B 轮质量原则检查，系统化发现、记录、修复问题，并沉淀可追溯的 `.bensz-api/task-{yyyymmdd-hhmm}-{简短描述}/auto-test-skill/output/plans/` 与 `.bensz-api/task-{yyyymmdd-hhmm}-{简短描述}/auto-test-skill/output/tests/` 文档。⚠️ 不适用：用户只是想优化功能（应直接修改）、只是询问技能问题（应直接回答）、没有明

当用户明确要求"使用 awesome-code / 多代理协作 / 并行协调开发"时使用。通过脚本收集可用 Agent 摘要、配置约束与 `dispatch_gate`，再由 AI 自主判断 single-pass / focused-agent / parallel / sequential 策略并选择子代理；当配置中的 required route agent 缺失时必须阻塞继续执行。⚠️ 不适用：用户仅需单一角色的简单修改或咨询、用户未明确表达多代理协作意图、用户只是了解技能概念。

当 Bensz 系列 skills 在真实用户环境中因 skill 设计缺陷而出现 bug，或用户明确说“我想 report bensz skills bugs”“帮我公开上报 bensz skills 的 bug”时使用。该 skill 负责把 bug 规范化记录到 `~/.bensz-skills/bugs/`，并在用户明确要求公开报告时通过本地 `gh` 轻量上传到 `huangwb8/bensz-bugs`，全程严禁修改用户本地 Claude Code/Codex 中已安装 skills 的源代码。

当用户明确要求"优化 prompt"、"改进提示词"、"润色指令"或"将简陋 prompt 转换为最佳实践版本"时使用。基于 OpenAI 和 Anthropic 官方最佳实践，对用户提供的简陋 prompt 进行结构化优化，输出符合社区标准的高质量版本。

当用户明确要求“压缩/瘦身/精简某个 Agent Skill 的 Markdown 文档”“在不改变功能前提下降低 skill 上下文开销”时使用。先理解目标 skill 的真实能力与安全边界，再在忽略 `tests/`、`plans/` 以及目标 skill 的 `README.md`、`CHANGELOG.md` 的前提下，压缩 `SKILL.md`、`references/*.md` 等工作型 Markdown，并把中间产物隔离到 `.bensz-api/task-{yyyymmdd-hhmm}-{简短描述}/compact-bensz-skill

当用户明确要求"提交 Git 改动"、"生成 commit 信息"或"创建 git commit"时使用。仅用 Git 分析改动并自动生成 conventional commit 信息（可选 emoji）；必要时建议拆分提交，默认运行本地 Git 钩子（可 --no-verify 跳过），提交后默认自动 push（可 --no-push 跳过）。

当用户明确要求“review 某个 GitHub PR”“评估某个 pull request 是否值得 merge”“帮我判断这个 PR 怎么处理”时使用。基于用户提供的 GitHub 仓库地址、PR 编号/链接和补充说明，进行只读、证据驱动的 PR 审查：理解 PR 解决的问题、评估方案优劣与局限、默认优先使用内置“好 PR”标准并在必要时联网补充、识别恶意或高风险改动，并输出是否建议 merge 的 Markdown 决策报告。⚠️ 不适用：用户要你直接修改 PR 代码、直接 merge PR、或在本地执行 PR 分支中的不可信代码。

当用户明确要求"发布项目到 GitHub"、"创建 GitHub Release"或"生成 Release Notes"时使用。智能分析 tag 间历史变化，生成专业且吸引人的 Release Notes，自动创建 GitHub Release。支持首次发布、常规版本、预发布版本（alpha/beta/rc），自动识别 prerelease 标记。

当用户明确要求"初始化项目"、"创建项目指令文件"或"生成 AGENTS.md"时使用。完全自动化：自动检测操作系统默认语言，分析项目目录结构（支持 Python/Web/Rust/Go/Java/数据科学/文档项目等），推断项目类型和用途，一键生成规范的项目指令文档。生成结果包括：AGENTS.md（跨平台通用项目指令，Single Source of Truth）、CLAUDE.md（Claude Code 特定适配，通过 @./AGENTS.md 引用）、README.md（项目介绍与使用方法）、CHANGELOG.md（项目变更记录）、.giti

当需要把本仓库 skills/alpha 下的生产 skills 安装到系统级（默认同时安装到 Codex: ~/.codex/skills 和 Claude Code: ~/.claude/skills），以便在任意项目/对话中可被发现与调用时使用。默认不安装 skills/beta；只有显式指定 beta 源目录时才处理 beta skill。使用 MD5 哈希进行版本控制，仅安装有更新的 skills；支持 --skill 指定单个或少量技能安装/更新、强制覆盖安装、指定单一目标安装和远程安装模式（--remote --check/--auto）。

当用户明确要求"并行执行同一条 Vibe Coding 指令 / 多个独立 agent 或 subagent 同时审查、想方案、优化、对比多条路线 / 多线程独立尝试"时使用。默认使用智能模式：由宿主原生 subagent 独立分析并由主 agent 汇总；智能模式和代码模式必须使用同一套 `.bensz-api/task-{yyyymmdd-hhmm}-{简短描述}/parallel-vibe/{yyyy-mm-dd-hh-mm}/` 运行目录、`@main/plan.json`、thread `workspace/`、`RESULT.md` 与 `r

当用户明确要求"转换图片格式"、"修改图片格式"、"图片格式转换"时使用。支持任意格式图片到目标格式的转换，包括：本地文件/网络URL/剪贴板图片输入，PNG/JPEG/WEBP 等常见格式输出，单文件或批量处理模式。核心特点：自动检测输入格式、支持透明度处理、批量处理保持原始文件名结构。⚠️ 不适用：用户只是想调整图片大小/裁剪（应使用图片编辑工具）、只是想查看图片信息（应直接使用文件查看器）、没有明确"格式转换"意图。

规范 AI 开发 R Markdown 分析脚本的行为准则。当用户要求"写 Rmd 分析"、"开发 R 脚本"、"做数据分析"时触发。核心原则：遵循主业与副业分离架构（.R 保留完整数据，.Rmd 应用业务阈值），优先使用用户已有 R 包资源；图表默认按 Nature 级别可读性与出版质量生成；专家级解读兼顾弱背景读者，提供四层框架、指标导读与不常用指标首次解释协议；路径验证确保跨平台兼容性。前提：luckyBase 为硬依赖。

Fourteen posts of being wrong in production, compressed to checkboxes

Healthy nodes, a quiet network, 300 restarts in three days, and a latency budget measured in milliseconds

Discovery worked. Ping worked. Every TCP connection timed out, and later the tunnel only worked when someone had a terminal open.

Every VM came back. The cluster did not. Declarative systems converge on config, and the datapath isn't config.

A surprising share of AI-in-the-terminal failures aren't the AI. They're zsh, and a version of bash from 2006.

A Claude Code plugin turns standalone project configuration into a namespaced, installable extension that teams and communities can update as one unit.

None of the safety came from the model. It came from six boring habits.

Skills package instructions and references. Subagents run work in a separate context and return results. They solve different problems and can be composed deliberately.

Six hours in, one step left, everything green, and the incident that didn't happen

CLAUDE.md carries persistent project context. Skills load reusable procedures when relevant. Separating stable facts from task-specific workflows keeps both easier to maintain.

Twenty minutes recovering secrets that never existed, and the one sentence from a human that ended it

An API request routing a model's tool call through an approval gate to a remote MCP server

31 config keys, two audits, and why the first one was wrong in both directions

The official MCP Registry stores standardized server metadata rather than package code. Publishers verify a namespace, describe installation or remote access, and submit immutable versions.

Everyone looks at the Dockerfile. The file that actually leaked the key was the project file.

Remote MCP authorization uses established OAuth standards, but secure integration still requires issuer validation, least-privilege scopes, protected token handling, and server-side enforcement.

"Copy it over and switch the reference" is two steps, and the outage lives in the one nobody checks

stdio fits local processes and prototypes. Streamable HTTP fits hosted services and shared integrations. The right choice follows where the capability runs and who must reach it.

The most important rule wasn't about what I could change. It was about what I was allowed to display.

Tools perform operations, resources expose readable context, and prompts provide reusable templates. Choosing the correct primitive makes an MCP server easier to understand and govern.

Run autonomous hunt loop on a target — scope check → recon → rank surface → hunt → validate → report with configurable checkpoints. Usage: /autopilot target.com [--paranoid|--normal|--yolo]

Build an exploit chain — given bug A, finds B and C to combine for higher severity and payout. Knows common chain patterns: IDOR→ATO, SSRF→cloud metadata, XSS→ATO, open redirect→OAuth theft, S3→bundle→secret→OAuth. Usage: /chain

Active vulnerability hunting. Two-track dispatcher — asks Red Team vs WAPT, hands off to hunt-dispatch skill and sibling commands. Usage: /hunt target.com | /hunt *.target.com | /hunt targets.txt [--vuln-class X] [--source-code P] [--chrome]

On-demand intelligence fetch for a target — CVEs, disclosed reports, new features. Pulls NVD/GitHub-Advisory CVEs + bundled disclosed reports + hunt memory context. Usage: /intel target.com

Inspect or rotate the autopilot ledger JSONL files (findings.jsonl, negatives.jsonl). Caps file size and keeps N rotated backups so memory does not grow unbounded.

Pick up a previous hunt on a target — shows hunt history and untested surface from the autopilot ledger. Usage: /pickup target.com

Run full recon pipeline on a target — subdomain enum (Chaos API + subfinder), live host discovery (dnsx + httpx), URL crawl (katana + waybackurls + gau), gf pattern classification, nuclei scan. Outputs to recon/<target>/ directory. Usage: /recon target.com

Optional manual note on a target or the last confirmed finding. Capture is automatic during autopilot; this is for extra context. Usage: /remember

Write a submission-ready bug bounty report. Generates H1/Bugcrowd/Intigriti/Immunefi format with CVSS 3.1 score, proof of concept, impact statement, and remediation. Run /validate first. Usage: /report

Mandatory pre-flight scope check — verify an asset is in scope BEFORE any HTTP touch. Deterministic (deny-wins, default-deny) via engine/scope.py against the engagement's scope.md. Blocks out-of-scope testing. Usage: /scope <asset> [<asset> ...]

Show ranked attack surface for a target from its recon manifest + hunt memory. Deterministic backing is `cbh surface <target>` (reads recon/<target>/manifest.json); LLM layer adds ledger signal. Usage: /surface target.com

Meme coin and token security scan — checks for rug pull vectors (hidden mint, honeypot, fee manipulation, LP lock bypass, authority retention, bonding curve exploits, fake renounce, sandwich amplification). Manual 8-class grep audit (with an optional automated scanner if present). Usage: /token-scan <contract_path_or_dir> [--chain solana]

Quick 7-Question Gate triage on a finding before writing a report. Kills N/A submissions before they happen. Faster than /validate — for quick go/no-go decisions. Usage: /triage

Validate a finding — runs 7-Question Gate + 4-gate checklist. Kills weak findings before report writing. Prevents N/A submissions that hurt validity ratio. Usage: /validate

Smart contract security audit — runs through 10 bug class checklist (accounting desync, access control, incomplete path, off-by-one, oracle errors, ERC4626, reentrancy, flash loan, signature replay, proxy/upgrade). Applies pre-dive kill signals first. Generates Foundry PoC template for confirmed findings. Usage: /web3-audit <contract.sol>

Crabbox is a single CLI (`crabbox`). Commands are top-level, not nested under a

`crabbox actions` prepares a leased box from your repository's own GitHub

See [Runtime adapter stack](../features/runtime-adapter-stack.md) for the

`crabbox admin` groups trusted operator controls for coordinator-backed leases and the cloud resources behind them. Use it to inspect every lease the broker tracks, reconcile expired leases against live cloud state, force-release or delete a backing server, print provider IAM pol

`crabbox artifacts` turns a desktop lease into durable QA evidence: it collects

Make any song you can imagine

Leading AI-powered video generation platform that specializes in creating hyper-realistic talking avatars

Hermes Agent is an open-source, self-improving autonomous AI agent developed by Nous Research

Kilo Code is a popular, open-source AI coding agent and "agentic engineering" platform designed to help developers build, refactor, and debug software faster

General-purpose agent in one static Go binary. ReAct loop, ACP server for IDEs, OpenAI-compatible REST API with embedded web UI, Telegram gateway, cron schedule…

Autonomous agent framework with structured memory, safety hooks, and loop management. Built by the agent that runs on it.

TSP自托管、零运维的 A 股「选股 + 监控 + 回测」量化工作台 | 基于 TickFlow 数据源 | LLM能力驱使策略定制+个股分析+复盘 | 自由接入第三方数据源与个性化扩展数据 | 个人开源 ,非TickFlow官方项目

Curated, verified Agent Skills powered by ModelStudio.

Run Claude Code, Codex, Antigravity, Cursor Agent and OpenCode as one runtime — persistent sessions, multi-agent councils, an OpenAI-compatible endpoint, an MCP…

pi had nothing (nothing), so I made something (something) — sorry mariozechner-senpai, I went ahead and lovingly soiled your pure pi for you. opinionated fork o…

A persistent workspace for development work that self-improves and continues beyond one session.

Open-source memory and context for user-aware agents: scoped memory, provenance, retrieval quality, correction, boundaries, evals, and MCP/HTTP access.

📚 A zero-dependency, git-backed micro-lesson library for AI Agents to asynchronously share and search verified debugging experience. Python stdlib only. | http…

Deterministic, local-first memory and guardrails for AI coding agents with no LLM in the hot path.

Deterministic spec-orchestration for local LLMs in the pi coding agent — drives prompts through refine→research→grill→compose→critique, with bundled web/docs/fe…

Native Safari browser automation for AI agents. 97 tools via AppleScript — zero overhead, keeps logins, runs silently in background. Drop-in alternative to Chro…

Agent OS: keep specialist agents in a hub, spin up a temporary orchestrator per task. Local-first, works with any model.

Git for agent memory. Branches, diffs, PRs, and rollback for what your agents know.

Multi-Provider AI Gateway - No personal logs by design. Model autodiscovery, Failover groups, High availability, Android companion app, and more - "Because we h…

Production-grade MCP server for MikroTik RouterOS with secure AI-native network automation.