qzruncode
UserAI is stupid and needs to be trained by humans.
Categories
Indexed Skills (17)
batch-code-scan
Use when the user invokes /human-plan:batch-code-scan or asks to scan the whole current project once, build a complete inventory of actionable code risks, split discovered problems into draft Human Plans, and either hand back the inventory or let loop mode start the first Plan without repeated rescans.
bug-fix
Use when the user invokes /human-plan:bug-fix after a concrete bug and its affected behavior have already been identified.
code-scan
Use when the user invokes /human-plan:code-scan or asks to find bugs, inefficiency, redundancy, maintainability problems, or other concrete code risks across the current project.
design
Use when the user invokes /human-plan:design or asks a senior product/UX/UI designer to create a decisive frontend experience brief for a page, workflow, state model, or usability-focused Human Plan before implementation and design-check review.
dev
Use when the user invokes /human-plan:dev with a requirement or an accepted Human Plan for a feature, behavior change, refactor, or other development work.
idea
Use when the user invokes /human-plan:idea or has a fuzzy thought, vague problem, product concern, or goal that is not yet a concrete requirement.
loop
Use when the user invokes /human-plan:loop to run, tick, inspect, or stop the Human Plan background batch loop.
cleanup
Systematically rescan, simplify, refactor, or rewrite the full product scope declared by docs/product-spec/feature-list.md, with mandatory file coverage and a persistent cleanup plan. Use when the user invokes /code-cleanup:cleanup, says repeated business iteration has created messy code, wants a scope-wide cleanup rather than one dead-file deletion, or needs duplication, complexity, oversized modules, inconsistent patterns, legacy paths, dependencies, and performance waste removed without losing any ready feature.
audit
审计整个仓库或指定范围内已经证实的问题与改进机会，覆盖正确性、安全、架构、可维护性、依赖、生态复用、性能、运维、UI/UX、无障碍和产品完整性。用户调用 /repo-guardian:audit，询问仓库有什么问题、要求代码库健康检查、寻找重复或过时方案，或者需要一份不修改实现的优先治理清单时使用。
feature-list
Initialize, maintain, update, or audit a complete code-free functional baseline at docs/product-spec/feature-list.md. Use when the user invokes /code-cleanup:feature-list, wants an exhaustive product feature list that any AI can use to reproduce the full product without seeing its code, provides an approved business change that must update the baseline, or asks whether the existing baseline is complete and current.
arch-check
Use when the user invokes /human-plan:arch-check or asks whether current business logic, product behavior, code, or architecture is reinventing wheels or diverging from mature solutions.
design-check
Use when the user invokes /human-plan:design-check or a proposed frontend change needs senior UX/UI review for product fit, task flow, interaction model, visual system, responsive behavior, and usability before implementation.
plan-check
Use when the user invokes /human-plan:plan-check or asks for architectural and code-impact review of a Human Plan before implementation.
worktree
Use when the user invokes /human-plan:worktree with a Plan Ref or needs a project-local git worktree paired one-to-one with a Human Plan so one plan can be executed in an isolated branch/session with the required local workflow context available.
align
Use when the user invokes /human-prompt:align or gives a short/vague task request that should become an aligned prompt before any AI starts work.
improve
调查并完成一个经过验证的仓库改进，可处理 repo-guardian 已发现的问题或用户直接指定的目标，包括缺陷、漏洞、架构债务、重复代码、过时依赖、性能瓶颈、UI/UX 与无障碍问题、功能缺失，以及用更好的现有方案替换自研实现。用户调用 /repo-guardian:improve，或要求实际修复、现代化、优化、升级、替换、移除、补齐仓库能力，而不是只要报告时使用。
roadmap
根据项目目标、当前实现、治理发现、用户与运维需求以及当前生态证据，决定仓库下一步应该做什么。用户调用 /repo-guardian:roadmap，询问项目下一步方向、功能优化或升级、缺失能力、现代化优先级、自研还是采用现有方案，或者需要有执行顺序的仓库演进路线图时使用。
Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.