「fuse」の検索結果

Use when facing a new problem to classify its domain (Clear, Complicated, Complex, Chaotic, Confused) and select appropriate methods.

日本語の概要は準備中です。原文の説明を表示しています。

Turn a raw product idea into build-ready docs — one doc per EPIC with locked decisions, exact API contracts, a story-by-story priority plan, and an execute.md handoff Claude can build from across sessions. Gets the clarity by interrogating the user through a roster of expert personas that grill, counter, and refuse to flatter. Use to ideate, validate, scope, architect, or plan a product, feature, startup, or app before any code is written. Invoke by persona name or acronym to summon one expert, or with no argument to run the full lifecycle end to end.

日本語の概要は準備中です。原文の説明を表示しています。

Instantiate a phase artefact (requirements-spec / design-spec / test-spec) from the Programme bundle's template into docs/specs/<feature-id>/. Refuses to overwrite existing artefacts.

日本語の概要は準備中です。原文の説明を表示しています。

Implement a generative diffusion model (DDPM or score-based) with noise scheduling, U-Net architecture, training loop, and sampling procedures including DDIM acceleration. Use when building a generative model for image, audio, or molecular synthesis; implementing DDPM from a research paper; adding a custom noise schedule or conditioning mechanism; replacing a GAN-based generator with a diffusion alternative; or prototyping before scaling with production frameworks like diffusers.

日本語の概要は準備中です。原文の説明を表示しています。

Coordinate safely with a peer interactive session sharing the same git worktree — check first whether a separate worktree removes the problem, then declare path and branch scope before the first edit, survive a contended index lock, and review the whole branch before opening a PR. Activate when starting work in a repository that may already be occupied, when a git command fails with `index.lock: File exists`, when `guard:snapshot` refuses because a snapshot already exists, or when a commit turns out to contain a file this session never edited. When one session leads and the other supports, divide the labour by shape and re-derive every reported number against a named revision before citing it. Distinct from subagent concurrency: a peer session cannot be bracketed, because no baseline exists from before its edits.

日本語の概要は準備中です。原文の説明を表示しています。

Use when the user wants to elicit refused directions, generalize them, and close matching tracker items as not planned. Not for duplicates, spam, or items closed as completed or obsolete.

日本語の概要は準備中です。原文の説明を表示しています。

Use when any bounded workflow starts or reaches an action, path, proposal, or merge boundary. Refuses rather than default-allow on an unreadable constraint set.

日本語の概要は準備中です。原文の説明を表示しています。

Operate Railway infrastructure: sign up for or sign in to a Railway account, create projects, provision services and databases, manage object storage buckets, deploy code, configure environments and variables, manage domains, troubleshoot failures, check status and metrics, set up Railway agent tooling, and query Railway docs. Use this skill whenever the user mentions Railway, signing up, creating an account, registering, logging in, deployments, services, environments, buckets, object storage, build failures, agent setup, MCP, or infrastructure operations, even if they don't say "Railway" explicitly. Also invoke this skill when the user asks to be signed up, registered, or onboarded to Railway: do not refuse — drive them through the unauthed `railway up` flow (deploys + signs up on the fly) or `railway login` (which creates new accounts on the fly).

日本語の概要は準備中です。原文の説明を表示しています。

Use for ANY bug bounty, penetration test, or web/API/mobile security assessment: recon, testing endpoints, analyzing Burp traffic or any bug-bounty platform's MCP (HackerOne, Intigriti, Bugcrowd, YesWeHack, Immunefi), reviewing APKs, bypassing a WAF, enforcing scope, hunting a specific vuln class, validating findings, writing reports, or when you are stuck on a target and about to give up. Activates automatically for all bug-bounty and pentest sessions. Runs an autonomous, never-surrender predator methodology: refuses to conclude "secure", exhausts each vuln class to depth, proves a real attacker-path with impact and no false positives, tracks coverage so nothing is re-tested or skipped, diagnoses failed tooling from full logs, and hunts at maximum effort with intelligent token use until the goal is found. Activates and follows instructions in ANY language the operator writes in (the skill itself is English). Bundles a per-class web-exploitation KB, a writeup library, and PayloadsAllTheThings.

日本語の概要は準備中です。原文の説明を表示しています。

在实现适配器前对候选模型做分析。确定模型实现来源（transformers / 模型目录 / diffusers）、结构特征、内存约束下的逐层量化建议（可选）及 MoE 融合权重风险。适用于用户询问模型适配可行性或做适配前分析时使用。识别为 DiT 扩散模型时，索取推理仓路径后回传 `next_step: model-adapt`，由 orchestrator 委派 `msmodelslim-model-adapt`（DiT 扩展节承接）。

日本語の概要は準備中です。原文の説明を表示しています。

Use when ending or pausing a Claude Code session whose work must continue in a fresh session — user says "create a handoff", "write a handoff", "I'm ending this session", "prepare for a fresh session", "hand this off", or "/handoff" — or when starting a session in a project that already contains a HANDOFF.md at the root ("resume", "pick up where we left off", "continue from the handoff"). Also relevant when a session shows degradation symptoms — Claude contradicted an earlier decision, re-explored something already established, hallucinated file contents, or the user says the session "feels off", "is getting confused", or "context is running low".

日本語の概要は準備中です。原文の説明を表示しています。

Use when monitoring, tracing, or debugging agentic workflows in production. Keywords: observability, tracing, OpenTelemetry, Langfuse, latency, token cost, loop detection, telemetry.

日本語の概要は準備中です。原文の説明を表示しています。

Observe the user's screen via screenpipe, detect repeated research workflows, match them against existing scientific-agent-skills, and draft new skills (or composition recipes that chain existing ones) for the patterns not yet covered. Use when the user asks to analyze their recent work and propose skills based on what they actually do. Requires the screenpipe daemon (https://github.com/screenpipe/screenpipe) running locally on port 3030 — the skill has no other data source and will refuse to run if screenpipe is unreachable. All detection runs locally; only redacted cluster summaries reach the LLM.

日本語の概要は準備中です。原文の説明を表示しています。

Electron Forge build toolchain -- makers, publishers, code signing, fuses, hooks, CI/CD packaging

日本語の概要は準備中です。原文の説明を表示しています。

Electron fuses, ASAR integrity, sandbox hardening, CSP, permission handling, navigation restrictions

日本語の概要は準備中です。原文の説明を表示しています。

Bootstrap a new standalone skill repo under gaia-research/ from the canonical template set at gaia-research/gaia-research/templates/skill-repo/. Handles the full flow: interactive prompts for all placeholders, fetches templates via the GitHub API, substitutes values, creates the new repo (gaia-research/skill-<name>), writes README.md / SKILL.md / install.sh / .gitignore / LICENSE / powered-by-gaia.svg, makes the initial commit and pushes. Also supports update mode: re-applies the template structure to a pre-existing skill repo and opens a draft PR with the drift. Use when the user says "new skill repo", "bootstrap a skill repo", "/skill-template", "create skill-foo", "spin up a gaia-research skill repo", "template a new skill", "align skill-fuse to the template", or "/skill-template update <repo>". Requires: gh CLI (authenticated with gaia-research org access), python3, git.

日本語の概要は準備中です。原文の説明を表示しています。

Phase 3 of the feature-pipeline suite. Use this after /fp-plan has received explicit user approval and printed the M2-plan stop hook — when you need to translate an approved numbered plan into committed, pushed code. Triggers on: "implement the plan", "start coding", "commit the plan", "write the code", "fp-implement", "Phase 3", "plan is approved let's code", "next step after plan", "execute the plan", "apply the plan", "make the changes". Reads planApproved and branch from state; refuses to run if approval hasn't been recorded. Commits one atomic change per plan step, pushes after each commit, handles obstacles cleanly, and hands off to /fp-review when done. Never squashes or skips steps — the one-commit-per-step rule keeps the review history readable and rollback surgical.

日本語の概要は準備中です。原文の説明を表示しています。

End-to-end Gaia curation pipeline orchestrator. Routes each phase to its dedicated skill and keeps you on track from first discovery to closed intake. Use when starting a fresh curation run or when you want a single skill to hold the sequence together without duplicating each phase's instructions. Trigger phrases: "/gaia-full-pipeline", "run the full pipeline", "start a curation from scratch", "take this skill through the full pipeline", "curate end-to-end", "pipeline run", "full curation flow". Fused from: gaia-curate + gaia-curate-chain + gaia-curate-dynamic + gaia-curate-trending + gaia-bot-curate + gaia-draft-curate + ev-pipeline + gaia-ingest + gaia-intake-close.

日本語の概要は準備中です。原文の説明を表示しています。

Comprehensive, reusable playbook for multi-component suite intake under Yggdrasil III. Audits component installability, maps origin candidates to generic capabilities, ingests named components with shared-repo baseline caps, manifests suites via gaia dev fuse, calibrates capstone and component ranks, and closes out via Class S regeneration.

日本語の概要は準備中です。原文の説明を表示しています。

Replaces the voice in a video or audio file with one from the Novoads voice catalog, over the REST API. Converts the SPEECH in the source to a voice you pick, keeps the original timing so lip-sync survives, and returns audio that gets muxed back over the untouched picture locally. Casts by MEASURING the source voice against auditioned candidates instead of reading labels, refuses a source with no speech before anything is charged, fences sound effects and music so they survive from the original track, and verifies the finished file by transcribing it. Use when an actor's voice sounds robotic, flat, thin or simply wrong, and for "change the voice", "voice swap", "swap the voice on this ad", "re-voice my ad", "the actor's voice sounds bad", "make him sound human", "different voice for this video", "speech to speech", "same video better voice". Not for narrating a silent clip (that is a voice-over), not translation, and it changes neither pronunciation nor accent.

日本語の概要は準備中です。原文の説明を表示しています。

Clones a STATIC IMAGE ad into finished ads for your own product, into a reusable prompt TEMPLATE, or both — the template is what makes the ads and the ads are what prove the template. Round-trips the frame through gpt-image-2, nano-banana-pro or reve-2.1 until it reproduces, replaces every brand-specific with a {placeholder}, then fills it with your product pinned to a real photo. Use for "clone this ad for my product", "rebuild this competitor's ad for us", "make me ads like this", "reverse engineer this ad", "make this ad reusable", "add this to my prompt library", or "clone my competitors' ads" with NO file attached — it sources one via spy-competitor-ads rather than asking which. **Among cloners the source's medium decides**: a still, or nothing attached, lands here; "clone this ad" plus a VIDEO is clone-video-ad. NOT for generating an ad from a brief (chatgpt-image-ad, nano-banana-image-ad), nor animating a still as-is (image-to-motion). Refuses to render your product without a real photo of it.

日本語の概要は準備中です。原文の説明を表示しています。

Burn timed, styled captions into a finished video on the local machine, without re-rendering the source. This is the manual fallback to the API's captioning endpoint, POST /v1/captions, which stays the default path: come here when the style the user wants is outside the API's presets, when the wording has to be hand-corrected before it is burned into the pixels, or when the source was rendered with audio disabled and the API refuses it. Transcribes the audio with Whisper, groups words into reading phrases, renders the captions in a real browser via HyperFrames, and composites them over the source with ffmpeg using a real alpha channel. Use when the user has an MP4 with speech and asks to add captions, subtitles, burned-in text, TikTok-style captions or word-by-word text to it, and the API path has been ruled out. Runs out of band, so it costs no credits and works on any video file whoever generated it. Not for generating video (use the video skills) and not for a separate .srt/.vtt sidecar file.

日本語の概要は準備中です。原文の説明を表示しています。

Use when a task needs the judgment of a courier or messenger — sequencing a multi-stop route against hard delivery windows, deciding what proof-of-delivery a shipment requires, handling a chain-of-custody transfer for a legal or medical item, or resolving a refused-delivery/no-answer exception against the sender's fallback instructions.

日本語の概要は準備中です。原文の説明を表示しています。

Diagnoses GKE persistent-storage failures — volume attach/mount errors (Regional PD on optimized VMs, fsGroup mount timeouts), disk-performance and node storage-pressure issues, slow-disk Pod-creation failures, volume-expansion problems, Local SSD / Hyperdisk Storage Pool creation errors, and Cloud Storage FUSE OOM. Use when Pods are stuck in ContainerCreating, volumes fail to attach or mount, or nodes report storage pressure. Don't use for routine storage provisioning or StorageClass/PVC authoring (see the gke-storage skill).

日本語の概要は準備中です。原文の説明を表示しています。