Explore

SQL injection — error-based, blind, union, time-based, WAF bypass. Must demonstrate data extraction or impact.

WebSocket, SSE, gRPC and GraphQL subscription authorization — handshake auth, per-channel and per-topic subscription authz, message replay, gRPC reflection and metadata trust, and the fact that these protocols bypass REST middleware entirely.

Race condition and concurrency flaws — TOCTOU, double-spend, double-redeem, limit bypass, and session races. Includes a controlled test harness and the reproducibility discipline needed to separate a real race from an app being slow. See hunt-business-logic for the state rules these break.

JS prototype pollution — __proto__, constructor, Object.assign, merge, clone. Chains to XSS, RCE, auth bypass.

Path traversal — directory traversal, file read, LFI, RFI, null byte, encoding bypass, ZIP/tar slip. Must demonstrate file read or impact.

Open redirect — parameter manipulation, path-based, host header, JavaScript location. Chains to OAuth abuse, phishing, ATO.

OAuth/SAML — redirect URI, JWT attacks, state bypass, token theft. Must demonstrate ATO or token theft.

Cross-session hunt memory and chain builder with state machine — session resume, lead board, pattern database, finding deduplication, bug chain discovery, and finding lifecycle management. Use when resuming a previous hunt session, tracking findings across sessions, or managing hunt state.

JWT attacks — none algorithm, alg confusion, key brute, claim manipulation, JKU/JWK injection, token substitution. Chains to auth bypass and ATO.

IDOR/BOLA — differential testing, two-session comparison, privilege escalation. Every candidate must pass 7 gates.

HTTP header injection — Host header, password reset poisoning, CRLF, response splitting, path bypass via headers.

GraphQL attack surface — introspection abuse, batching/rate-limit bypass, alias IDOR, directive injection, subscription DoS, field-suggestion info disclosure, schema extraction, injection-to-RCE.

HTTP request smuggling — CL.TE, TE.CL, TE.TE, H2.CL, prefix injection, request tunneling. Chains to credential hijacking, XSS, cache poisoning. Real bounty examples, working scripts, automation.

Insecure deserialization — PHP, Java, Python, Node.js, .NET, Ruby. Object injection, gadget chains, RCE.

CSRF — state-changing actions without CSRF tokens, SameSite bypass, subdomain CSRF, double-submit cookie, token prediction. Chains to ATO, RCE, privilege escalation.

CRLF injection — response splitting, header injection, XSS, cache poisoning, cookie injection. Must demonstrate concrete impact.

CORS — origin reflection, null origin, preflight abuse, wildcard detection. Establishes whether credentialed cross-origin reads are actually possible.

HTTP cache poisoning — unkeyed headers, fat GET, file extension confusion, web cache deception, exploit chains, and automation.

Business logic flaws — payment and price manipulation, coupon and credit abuse, referral and reward stacking, refund and chargeback paths, inventory and quantity limits, currency rounding, entitlement bypass. Highest severity per finding, entirely missed by scanners because no payload triggers them.

API key/secret detection — JS bundles, source code, config files, error messages, backup files. Must demonstrate sensitive capability impact.

API security — BOLA/IDOR, mass assignment, rate limit bypass, auth bypass, method confusion. Must pass 7 gates.

Evidence management — Structured proof artifacts, HTTP request/response capture, screenshots, tool output storage, and evidence-to-finding linking. Use when collecting evidence, capturing proof, organizing test results, or linking evidence to findings.

Password spray methodology for bug bounty — when to do it vs web-vuln hunting, the wordlist-gen + breach-check + osint-employees + spray pipeline, mode selection (http-form / oauth / o365 / okta), rate-limit + lockout tactics, legal guardrails, success detection, and the spray → authenticated /hunt chain pattern.

Treats every security control as a falsifiable claim rather than a fact — the 3x3 actor/resource matrix, per-class control ledger, and differential tests for authn, authz, tenant isolation, validation, encoding, rate limits, session and cache controls. Use when testing an app that appears to have protections, when a control looks like it works, or before concluding an area is clean.

Cloud and platform configuration assessment — object storage public access, over-permissive IAM, instance metadata reachability, exposed Kubernetes and container registry APIs, managed identity trust, and secret material in the open. Read-only assessment methods; see IMPACT BOUNDARY.

Maps vulnerability primitives to what they unlock, so a small finding reprioritizes the rest of the hunt and the report describes real escalation accurately. Two uses — choosing the next test, and writing the impact section. Use when you find something that looks minor but implies the stack is exposed, or when writing an impact statement that needs to convey severity honestly.

Framework-specific attack playbooks, mobile app testing, CI/CD pipeline attacks, HTTP request smuggling, cache poisoning, and MFA/2FA bypass patterns. With differential testing integrated. Use when targeting specific frameworks or need platform-specific attack vectors.

Real-world security assessment methodology with program-specific workflows, time management, and reporting discipline. Structured for accurate validated findings and honest coverage.

Comprehensive backend development methodology covering API design, authentication, database design, caching, testing, deployment, and monitoring. Full lifecycle from requirements through production.

Review system architecture: module boundaries, dependency direction, coupling analysis, and scalability assessment.

Use when adding or updating a benchmark adapter in the Exgentic repository. Follow the repository benchmark principles, keep the benchmark contract protocol-agnostic, prefer the thinnest possible wrapper that makes the benchmark accessible to any Exgentic agent, reuse external harness assets and scoring where possible, and validate the adapter with representative smoke tests before finishing.

Use when adding or updating an agent adapter in the Exgentic repository. Follow the repository agent principles, separate lightweight config from heavy execution logic, isolate third-party dependencies behind lazy imports, adapt to any benchmark contract without requiring benchmark changes, and validate the adapter with representative smoke tests before finishing.

Activate MUE — self-evolving AI agent. Type /mue to become MUE.

Use when writing, moving, reviewing, or auditing documentation in the deepseek-harness repo — choosing hierarchy and detail, separating tutorials from references, checking tutorial progression, trimming doc slop, responding to a verify-doc-budgets failure, or requests like "improve the docs", "audit the docs", "where should this be documented", or "this doc is too long".

Use when publishing, updating, moving, or removing DeepSeek Harness documentation website pages; editing website/docs.ts mappings or navigation; diagnosing a page missing from the VitePress site; fixing projected documentation links; or running the docs:dev, docs:check, and doc-sync workflow after website-content changes.

Use KMSG to read, watch, and send KakaoTalk messages safely from a coding agent. Use whenever the user mentions KakaoTalk or kmsg and wants chat discovery, recent messages, monitoring, text or image delivery, setup checks, or CLI troubleshooting. Do not use for other messaging platforms or private KakaoTalk protocol work.

Write durable work items with scoped descriptions, testable categorized acceptance criteria, governing references, and risk-matched guards

Maintain RFC and ADR artifacts without implementation work

Write precise, implementation-independent RFCs with testable normative clauses and correct lifecycle handoff

Execute a trivial non-behavioral change without unnecessary governance ceremony

Define reusable, non-interactive verification guards with stable commands and risk-matched scope

Execute governed implementation with work-item traceability, RFC/ADR authority, risk-scoped verification, and closure

Explore a governance design, resolve ambiguity, and draft RFC or ADR artifacts without implementation

Remove project-local govctl integration through an ownership-aware dry run, archival plan, explicit confirmation, and recoverable edits

Stress-test consequential design options with premortem and backcast reasoning, then return a risk-calibrated recommendation to the owning governance workflow

Write decision records that explain context, evaluated alternatives, chosen direction, and consequences without creating product obligations

Use when the user wants a disciplined software development workflow with design-first planning, implementation plans, TDD, systematic debugging, code review, or verification-before-completion, adapted from obra/superpowers.

Use when the user wants to publish or update a local skill in the NextClaw marketplace, especially when they need clear guidance for version readiness, platform login, username setup, scoped package naming, and post-publish verification.

Use when the user wants to use websites, browser login sessions, Electron apps, or external CLIs through a local OpenCLI setup, especially when setup guidance, readiness checks, and safe task execution are needed.

Use mmx to generate text, images, video, speech, and music via the MiniMax AI platform. Use when the user wants to create media content, chat with MiniMax models, perform web search, or manage MiniMax API resources from the terminal.

Use when the user wants to list, view, start, create, or update Linear issues from the terminal via schpet/linear-cli, including setup, auth, repo config, and safe read/write boundaries.

Use when the user wants to operate Lark or Feishu via the local lark-cli (@larksuite/cli), including install, app credentials, OAuth, readiness checks, and safe read/write boundaries.

Use when the user wants to publish an already-built static website or front-end app through GemiGo, such as a Vite/React/Vue static build, plain HTML/CSS/JS page, landing page, demo, docs site, or small browser app, and get a hosted public URL.

当用户希望把 Codex 或 Codex CLI/SDK 作为 NextClaw 正式会话类型接入，或要求安装、更新/升级 Codex NARP runtime、修复、doctor、排查版本或真实冒烟时使用。

Use when the user wants to search external ClawHub skills or install a ClawHub skill into a NextClaw workspace's skills directory without treating ClawHub as the NextClaw marketplace.

当用户希望把 Claude Code 作为 NextClaw 正式会话类型接入，尤其是通过 narp-stdio 配置、安装、修复、doctor、补齐 Runtime default 模型选项或真实冒烟时使用。

Use the local browser-connector CLI to inspect and operate the user's current Chrome tabs through the Browser Connector extension and Native Host. Use when the user asks to list open browser pages, read a current page, capture a screenshot, or perform controlled browser interactions from NextClaw.

Use the local aigen CLI to generate images through configured providers such as OpenRouter or OpenAI. Use when the user asks to create, generate, or render images and wants files produced locally through an external CLI.

使用 NextClaw 反馈工作流 CLI 读取、领取、处理和回评已批准的反馈；适用于处理端或被唤醒的 Codex，不用于管理员审批。

处理 NextClaw 私有讨论事件，核对参与者身份与反馈审批，按宿主输出模式持续协作。