Menu
SkillsMP a collecté 40 skills depuis JZKK720/cubecloud-agentic-os. Ouvrez un skill pour examiner sa source et ses détails.
Affichage de 40 skills collectés sur 40.
Use when the user wants to autonomously tune a numeric metric against a single editable file, with no human check-ins. Triggers: "autoresearch", "overnight tuning", "agent runs experiments", "improve val_* metric", "optimize loss", "benchmark my model",…
Langue du texte source : anglais
Use when starting a new project, reviewing code, refactoring, or onboarding a new contributor in TypeScript / JavaScript / React / Node.js / adjacent stacks. Triggers: "best practices", "conventions", "code style", "what's the right way to do this", "lint…
Langue du texte source : anglais
Use when writing a new skill, reviewing a skill, refactoring an existing one, or onboarding someone to the skill ecosystem. Triggers: "how do I write a good skill", "review this skill", "skill format", "skill frontmatter", "skill description contract", "skill…
Langue du texte source : anglais
Use when the user wants to create, build, fork, or find a skill for a workflow, or when a "new" skill idea already feels familiar. Triggers: "does a skill for X already exist", "find me a skill for…", "search skills", "is there a marketplace skill for…",…
Langue du texte source : anglais
Autonomous high-reasoning operating mode for long-running, multi-step work. Use when the user says "use fable mode", "plan end to end", "autonomous run", "long horizon", "self verify", or starts ambitious work meant to run start-to-finish with minimal…
Langue du texte source : anglais
Use when the agent wants to consolidate what was learned, decided, deferred, and what to skill-ify at the end of a long agentic task. Triggers: "wrap up", "summarise the session", "what did we do today", "end-of-task", "post-work", "structured retro", "I want…
Langue du texte source : anglais
Use when the user wants to create, build, structure, or templatize a skill, when a workflow is being repeated, or when the agent is asked "should this be a skill?". Triggers: "should this be a skill", "is this workflow skill-worthy", "write a skill properly",…
Langue du texte source : anglais
Use when the user reports a vague issue, a recurring symptom, a mystery behaviour, or asks "why is this happening?" — a problem with incomplete information, not a known bug. Triggers: "why is this happening", "mystery behaviour", "investigate this",…
Langue du texte source : anglais
Use when the user wants a plan stress-tested from a CEO/operator lens — strategy, scope, ROI, risk, and "is this the right thing to build?". Triggers: "review like a CEO", "pressure-test the plan", "is this worth doing", "what's the user-facing outcome",…
Langue du texte source : anglais
Use when reviewing a wireframe, mockup, prototype, or finished UI; when launching a new surface; or when the user asks "is this design good?". Triggers: "review the UX", "is this design good", "pressure-test the mockup", "is the 200ms hook right", "is the…
Langue du texte source : anglais
Use when a plan needs technical stress-testing, when sequencing is unclear, when the user asks for a tech lead's view, or when dependencies may block the build. Triggers: "tech lead review", "is this buildable", "stress-test the architecture", "what's the…
Langue du texte source : anglais
Use when the user wants a release-readiness review, when a feature is "done" but not yet tested, or when running pre-merge checks. Triggers: "is this safe to ship", "QA before release", "pre-merge checks", "is the build green from clean", "do the tests catch…
Langue du texte source : anglais
Use when the team needs to align on what worked, or when the user says "retro", "what did we learn", "post-mortem" — at the end of a project, sprint, or feature, before the next kickoff. Triggers: "retro", "post-mortem", "what did we learn", "what would we do…
Langue du texte source : anglais
Use when the user wants the agent to respond in ultra-compressed, low-token, no-filler prose while preserving technical accuracy. Triggers: "caveman mode", "talk like caveman", "use caveman", "less tokens", "be brief", "be terse", "drop the filler", "maximum…
Langue du texte source : anglais
Use when the user reports a bug, says "diagnose this", "debug this", "this is broken", "this is throwing", "this is failing", or describes a performance regression. Triggers: hard bug, flaky test, mystery behaviour, slow query, memory leak, intermittent…
Langue du texte source : anglais
Use when the user wants to stress-test a plan against the project's existing domain model, sharpen fuzzy terminology, and inline-update CONTEXT.md / ADRs as decisions crystallise. Triggers: "grill this plan", "stress-test the design", "challenge my…
Langue du texte source : anglais
Use when the user wants to find refactor opportunities that turn shallow modules into deep ones — deleting code rather than adding it. Triggers: "improve architecture", "find refactor opportunities", "deepen the modules", "this module is too shallow", "this…
Langue du texte source : anglais
Use when writing or modifying code, and the user wants test-first, red-green-refactor discipline. Triggers: "TDD this", "red-green-refactor", "test-first", "vertical slice", "write the test first", "behaviour test not implementation test", "no production code…
Langue du texte source : anglais
Use when the user wants to capture the current conversation as a PRD and publish it to the project issue tracker. Triggers: "create a PRD", "write a spec", "requirements doc", "capture this as a structured doc", "I want the next agent to pick this up", "make…
Langue du texte source : anglais
Use when the user wants to create, write, or build a new SKILL.md; when scaffolding a workflow into a reusable skill; or when an existing skill needs cleanup. Triggers: "write a skill", "create a skill", "SKILL.md template", "skill structure", "skill…
Langue du texte source : anglais
Use when starting any creative work — creating features, building components, adding functionality, modifying behaviour, or designing a system. Activates before any implementation, including TDD. Socratic design refinement: asks questions one at a time, then…
Langue du texte source : anglais
Use when something is broken, flaky, or behaving unexpectedly — enforces a 4-phase root-cause process: reproduce → hypothesise → instrument → fix with regression test. Replaces ad-hoc "change things until it works" debugging.
Langue du texte source : anglais
Use when a plan exists and is approved — runs the plan task-by-task with checkpoint reviews. Each task is a fresh subagent invocation that returns a clean diff and verification output.
Langue du texte source : anglais
Use when the plan is executed and the tests pass — verifies the worktree is ready, presents the user with the merge/PR/keep/discard options, and cleans up the worktree. The last skill in the development lifecycle.
Langue du texte source : anglais
Use when you need to fan out one-off work in parallel — research, multi-file exploration, search-across-many, or independent investigations. Distinct from `cubecloud-subagents` (which executes a plan); this is for ad-hoc parallel queries.
Langue du texte source : anglais
Use when the design is approved — breaks the work into bite-sized tasks (2–5 minutes each), with exact file paths, complete code, and verification steps for each task. Produces a plan a junior engineer with no taste and no context can follow.
Langue du texte source : anglais
Use when responding to code-review feedback — separates signal from noise, acknowledges what's right, defends with reasoning when appropriate, fixes what's wrong, and never accepts drive-by refactors in the same PR.
Langue du texte source : anglais
Use when about to hand work to a reviewer (human or agent) — runs a pre-review checklist to catch the obvious issues, reports findings by severity, and explicitly blocks on critical issues. Saves the reviewer's time and the author's embarrassment.
Langue du texte source : anglais
Use when starting any conversation or before any response — establishes that the agent must check for relevant skills before answering, asking clarifying questions, or taking action. Replaces ad-hoc reasoning with skill-first reasoning.
Langue du texte source : anglais
Use when the plan has independent tasks and the work can be parallelised — dispatches a fresh subagent per task with two-stage review (spec compliance, then code quality). Faster than sequential execution for non-blocking work.
Langue du texte source : anglais
Use when writing or modifying any code — enforces the RED-GREEN-REFACTOR cycle: write a failing test, watch it fail, write minimal code, watch it pass, commit. Deletes code written before tests.
Langue du texte source : anglais
Use when declaring a fix complete, a feature shipped, or a task done — enforces evidence over claims: did you actually run it, actually see the green test, actually see the user-facing behaviour? Replaces "it should work" with "I verified it works".
Langue du texte source : anglais
Use when the design is approved, before implementation begins, to create an isolated git worktree on a new branch and verify a clean test baseline. Triggers: "start a feature", "isolated branch", "before implementation", "set up a worktree", "I want to keep…
Langue du texte source : anglais
Use when creating new skills, editing existing skills, or verifying skills work before deployment — applies the TDD-for-skills approach: red phase (run the agent without the skill, capture failures), green phase (write the minimal skill that fixes the…
Langue du texte source : anglais
Full audit and smoke-test workflow for the Cubecloud Agent Desktop runtimes, sub-runtimes, IPC surface, CI gate, and asar integrity. Use when the task mentions "audit agent-desktop", "smoke test runtimes", "verify CI is real", "full audit before release",…
Langue du texte source : anglais
Anti-slop renderer design rules for agent-desktop surfaces (Memory tabs, Plans, Skills, Settings, modals, dialogs). Adapted from upstream taste-skill v2 - brief inference, three dials (VARIANCE / MOTION / DENSITY), AI-tell bans, strict pre-flight check.
Langue du texte source : anglais
Behavioral guidelines to reduce common LLM coding mistakes - think before coding, simplicity first, surgical changes, goal-driven execution. Use when writing, reviewing, or refactoring code in Agent Desktop work.
Langue du texte source : anglais
Handles Headroom context-compression workflows for Cubecloud desktop, Copilot, and VS Code. Use when the task mentions headroom, context compression, token savings, large tool outputs, long logs, CodeGraph bundle compression, headroom mcp, headroom proxy,…
Langue du texte source : anglais
Handles Cubecloud monorepo doc sync, translation inventory updates, README PDF rerendering, and screenshot refresh sequencing. Use when working on README.i18n.md, zh-CN/ja-JP/ko-KR doc translations, sync-docs.ps1 hardlink behavior, combined README PDF…
Langue du texte source : anglais
Use when writing, reviewing, or refactoring code, and the user wants the four principles applied — Think Before Coding, Simplicity First, Surgical Changes, Goal-Driven Execution. Triggers: "follow the four principles", "be surgical", "simplify first", "stop…
Langue du texte source : anglais