Ninthless

@ninthless source repo

13 skills
0 followers
17 hours ago last updated

▌
Write API Docs · ninthless bundle
Create, generate, write, update, rewrite, reconcile, or review an API document or contract from routes, handlers, schemas, types, client calls, tests, runtime evidence, OpenAPI/Swagger, GraphQL, gRPC, webhooks, AsyncAPI, or messages. Use when the final artifact is an API reference, endpoint documentation, API specification, frontend-to-backend or backend-to-frontend handoff, integration guide, or API documentation review. Trigger Chinese requests such as 接口文档, API文档, 接口说明, 接口契约, 接口规范, 接口对接文档, 根据前端生成接口文档, 根据后端代码补齐接口文档, and documentation work phrased as 整理, 补齐, 重写, 对接, or review. Document only evidence-supported behavior; never invent fields, constraints, permissions, errors, side effects, or deployment status. Do not trigger for endpoint/client/SDK implementation, webpage changes, tests, ordinary source review, bug diagnosis, generic HTTP teaching, API research without a documentation deliverable, translation, or incidental API mentions.

0
installs
▌
Websearch First · ninthless bundle
Use this skill by default for every user request: search the web before answering, explaining, planning, reviewing, coding, editing local files, running commands, or making recommendations. Search authoritative sources for current facts, unfamiliar domains, APIs, libraries, versions, standards, policies, security, prices, platform behavior, and implementation choices, and still perform a brief relevance search for simple local tasks. Trigger for 先联网搜索, 先查资料, 先搜索, web search, browse first, do research first, or webserach. Reconcile web evidence with local repository, runtime, and user-provided facts; cite sources that affect the result. Skip only when the user explicitly forbids network access or the search capability is unavailable, and report the exception.

0
installs
▌
No Code Comments · ninthless bundle
Keep generated or modified code-like artifacts comment-free unless the user explicitly requests comments or tooling requires a directive. Use as a companion whenever a task writes, edits, refactors, scaffolds, patches, regenerates, or shows source, configs, tests, scripts, snippets, migrations, schemas, workflows, or Dockerfiles. Trigger for clean code, production-ready code, 不要注释, 无注释, or 干净代码. Do not use for 写好注释, 优雅注释, JSDoc/docstring authoring, when quality-code-comments is the installed comment policy, or for pure analysis, review, planning, or explanation when no code-like content will be written or shown.

0
installs
▌
Xposed Module Dev · ninthless bundle
Create, review, migrate, or debug Android Xposed and LSPosed modules. Use for libxposed versus legacy XposedBridge decisions, method or resource hooks, app and system_server hooks, scope.list, module.prop, xposed_init, LSPosed activation, module-not-showing issues, hook-related logcat crashes, and 写 Xposed 模块. Trigger when the task depends on Xposed-compatible runtime hooking or module packaging. Do not use for ordinary Android, Gradle, Magisk, root, APK, or Java/Kotlin work that does not involve Xposed or LSPosed.

0
installs
▌
English Spec First · ninthless bundle
Normalize rough, mixed-language, or materially ambiguous requirements into a concise English working spec before planning or implementation. Use when the user asks to think in English first, 先用英文, 整理需求, professionalize a prompt, clarify scope, or turn messy product or coding notes into an actionable brief. Do not use for simple translation or proofreading, summarizing an already-final specification, reviewing a plan without rewriting it, or directly implementing a request whose scope is already clear.

0
installs
▌
Git Checkpoint Push · ninthless bundle
Safely stage, commit, checkpoint, or push completed work using the full Conventional Commits 1.0.0 structure: type, optional scope, optional breaking marker, description, meaningful body when context is not obvious, and valid footer trailers when applicable. Use only when the user explicitly asks to save work in git or GitHub, 提交代码, 推送, 保存进度, Conventional Commit, commit message, milestone push, or GitHub PR handoff. Inspect status, remotes, upstream divergence, verification, unrelated changes, compatibility impact, issue references, and attribution before writing history. Do not trigger for read-only git status, diff, log, review, conceptual Git questions, or imaginary commit examples that do not authorize repository mutation.

0
installs
▌
Build User Facing UI · ninthless bundle
Build, redesign, fully refactor, implement, review, or fix user-facing UI for web, dashboards, commerce, mobile, desktop/Electron, games, editors, kiosks, TV, wearables, and spatial interfaces. Use for hierarchy, layout, responsiveness, interactions, accessibility, platform adaptation, design systems, trust UX, performance, content design, Figma or screenshot implementation, and polished, distinctive, usable UI. Trigger for 重构界面, 重构前端, 重构所有 UI, 彻底重做 UI, rebuild the entire UI, and visual overhaul. For broad UI refactors, perform greenfield reconstruction followed by migration: extract product contracts, quarantine the old visual system, and design a new product-specific shell, navigation, layouts, tokens, components, and responsive system from a blank canvas. Do not visually derive the replacement from the old UI unless explicitly asked to preserve part of it. Exclude backend/API/database/CLI work, standalone image or logo creation, translation-only work, policy prose, backend performance, and headless logic.

0
installs
▌
Learn While Building · ninthless bundle
Mandatory passive-learning companion whenever the agent implements, debugs, refactors, or explains project code, including ordinary vibecoding. Always trigger for write, create, implement, build, fix, edit, modify, refactor, scaffold, 帮我写, 实现, 修一下, 改代码, vibe coding, and similar delivery work so the user absorbs the relevant model, decisions, and verification without asking. Default to passive teaching: finish the work first, then a compact learning overlay; do not quiz, withhold code, or ask what to learn. Upgrade to guided or practice only for 边做边学, 教我, 带我理解, learning mode, or predict-first requests. Do not trigger for result-only requests that forbid explanation, generic concept questions unrelated to current project work, git-only status/commit/push, or pure review/status with no code walkthrough.

0
installs
▌
Quality Code Comments · ninthless bundle
Mandatory comment-quality companion whenever the agent will generate, show, edit, patch, refactor, format, scaffold, or otherwise produce code or a code-like artifact. Always trigger before writing source, snippets, scripts, shell commands, queries, regex, configs, schemas, migrations, tests, fixtures, workflows, Dockerfiles, pseudocode, examples, or proposed patches, including trivial, throwaway, production, and explicitly comment-free requests. Trigger for write, create, generate, implement, fix, modify, refactor, 写代码, 帮我写, 实现, 修一下, 改代码, 重构, 加注释, JSDoc, docstring, and similar delivery work. It governs whether each comment should exist: prefer clear code, preserve required directives and documentation contracts, and add only accurate information the artifact cannot express locally. A no-comments request means zero optional comments; it never disables this skill. Do not trigger for read-only explanation, planning, inspection, diagnosis, status, or review when no code-like content will be produced.

0
installs
▌
Freelance Order Triage · ninthless bundle
Evaluate incoming freelance or client coding work before acceptance, quoting, or implementation. Use when deciding whether to take, clarify, reshape, stage, price, or decline a commercial job, student request, small change, bugfix order, rescue project, or AI-generated code cleanup. Trigger for client-facing questions, hidden-scope analysis, quote posture, revision limits, milestones, and acceptance criteria. Do not use for ordinary coding or debugging merely because it involves a bugfix or existing project unless the request is framed as an order, client engagement, or delivery decision.

0
installs
▌
High Constraint Coding · ninthless bundle
Mandatory coding-quality and maintainable-architecture companion whenever the agent will generate, write, create, implement, edit, modify, patch, refactor, convert, format, scaffold, regenerate, or otherwise output code or a code-like artifact. Always trigger before producing or changing source, snippets, scripts, commands, queries, regex, configs, schemas, migrations, tests, fixtures, generated code, build files, CI workflows, API definitions, pseudocode, examples, or proposed patches, whether in a repository or only in chat and whether trivial, throwaway, hypothetical, or production. Compose with specialized UI, API, security, platform, or documentation skills. Preserve contracts and ownership, make the smallest maintainable result, and verify proportionately. Do not trigger for explanation, planning, inspection, diagnosis, status, or review when no code-like output will be produced or modified.

0
installs
▌
Vibecoding Domain Scout · ninthless bundle
Research unfamiliar, current, regulated, or platform-dependent domains before product scoping or coding. Use when an app, product, or tool idea depends on industry workflows, business rules, policies, standards, compliance, real-world operations, or expert terminology the user or agent may not know. Trigger for domain maps, hidden requirements, workflow discovery, MVP cuts, risk and validation criteria, and build-ready briefs grounded in current sources. Do not use for routine apps in familiar domains, pure UI implementation, generic market summaries, or clear technical tasks that do not require domain discovery.

0
installs
▌
Powershell Safe Commands · ninthless bundle
Author or repair PowerShell commands safely on Windows when shell parsing can change meaning. Use for $, $_, $env:, @{u}, $(...), nested powershell -Command, here-strings, regex replacements, quoted paths, file edits from shell, generated one-liners, or failures such as $ 被吃了, PowerShell 外层吃了, unterminated strings, and ExpectedValueExpression. Trigger before rerunning a command affected by interpolation, quoting, or wrapper layers, and prefer direct executables, Python, or temporary scripts when safer. Do not use for simple direct commands with no PowerShell-specific syntax, conceptual explanations only, or non-Windows shells.

0
installs

13 published skills