134 skills · sorted by relevance
Manages context passing between DAG nodes and spawned agents. Handles context summarization, selective forwarding, and token budget optimization. Activate on 'bridge context', 'pass context', 'summarize context', 'context management', 'agent context'. NOT for execution (use dag-parallel-executor) or aggregation (use dag-result-aggregator).
The English summary is being prepared. Showing the original description.
Coordinate multiple agents to maintain context-mode, from sorting GitHub issues and reviewing pull requests to preparing releases and announcements. Workflows require reproduced bugs, source evidence, test-first changes, and checks across supported platforms.
Configure large language model training with NVIDIA Megatron-Core across multiple GPUs. This skill helps choose how to split computation, set training parameters, monitor GPU efficiency, and address memory limits or unstable training.
Design workflows in which AI agents handle multiple steps or specialist tasks. Choose a suitable structure, generate starter configurations, and define information handoffs, retries, validation checks, and limits on cost and context size.
Design systems where multiple agents work with separate task information. Choose coordination structures, define handoffs and shared state, and plan parallel work while checking whether communication costs and failure risks justify using more than one agent.
Configure large language model training across multiple NVIDIA GPUs with Megatron-Core. The skill helps choose how to split models and workloads, set training parameters, monitor performance, and address memory shortages or low GPU utilization.
親イシュー配下のサブイシュー（孫以下の全子孫を含む）を依存順を保ちつつ worktree で並列に自動実装・push 前 review・PR 作成・CI 監視・マージ可能状態化まで一括自動化。 「イシューツリーを並列実装」「配下のサブイシューをまとめて実装」「ツリー全体を並列で実装して」「イシュー階層を自動開発」で使用。 per-issue 計画立案（Plan: セッション継承モデル）→実装（Implement: sonnet）の分業。push 前 review（Review 通過後にのみ push・PR 作成して CI を 1 回だけ起動）。 外部チェック構成は args の externalChecks で明示（{"app", "context"} の組で宣言。[] で「なし」を確定して不要待機なし・未指定なら自動マージ停止・slug のみの旧形式は自動マージ fail-closed）。 自動 squash merge は autoMerge: true + externalChecks 明示（全 App の信頼済み context 宣言込み）の opt-in で実行（merge-exec の自己取得再検証 + サーバー側 branch protection 実測が前提。既定 false はマージ可能状態で停止し人間がマージ。サーバー側 workflow サンプル（upstream の docs/implement-issue-tree/auto-merge-sample.yml 参照）+ branch protection への委譲も可）。並列度（parallel）と依存（dependsOn）で実行順を制御。 Phase 単位で厳密に直列化したい場合は phaseGate: true（opt-in。ルート直下の子〔Phase 親〕を sub-issues リスト順に直列化し、前 Phase の全子孫が merged/closed になるまで次 Phase 配下に着手しない。既定 false は現行の post-order 優先度のみ）。 着手できる件数・blocked になる件数を事前に知りたい場合は dryRun: true（何も書かずに見積りだけ返す）。 単一イシューの実装は implement-issue、PR レビューは implement-review-pr を参照。
The English summary is being prepared. Showing the original description.
Turn a large engineering objective into a written plan with steps sized for individual pull requests, or proposed code changes. Each step includes context, dependencies, checks, and completion criteria so another agent can continue the work.
Configure large language model pretraining with torchtitan across multiple GPUs. Covers ways to split training work, launch jobs on computing clusters, enable Float8 numerical processing, and save or resume training through distributed checkpoints.
Delegate coding, analysis, planning, and reviews from Codex to Claude Code through its command-line interface. Capture responses, resume sessions, or run independent tasks in parallel, then verify changed files and test results independently.
Write and review launch prompts for agents working autonomously for hours or days. Define successful deliverables, excluded partial results, stopping rules, worker handoffs, and evidence-based review requirements so completion can be assessed against explicit criteria.
Configure large language model pretraining with torchtitan across multiple GPUs. Covers ways to split training work, launch jobs on computing clusters, enable Float8 numerical processing, and save or resume training through distributed checkpoints.
Turn feature requirements into a PRD, a document defining what to build, then organize development plans and GitHub Issues. Coordinate parallel agents and track progress, blockers, and next priorities through delivery.
Decomposes a contextualized plan into atomic, independently executable tasks with complete embedded context, registered through the plan API. Use after SAM Stage 3 Context Integration produces the contextualized plan artifact — when the plan is ready for task generation with CLEAR ordering, CoVe checks, and dependency graphs for parallel execution.
The English summary is being prepared. Showing the original description.
Guide for optimizing MSBuild build parallelism and multi-project scheduling. Only activate in MSBuild/.NET build context. USE FOR: builds not utilizing all CPU cores, speeding up multi-project solutions, evaluating graph build mode (/graph), build time not improving with -m flag, understanding project dependency topology. Note: /maxcpucount default is 1 (sequential) — always use -m for parallel builds. Covers /maxcpucount, graph build for better scheduling and isolation, BuildInParallel on MSBuild task, reducing unnecessary ProjectReferences, solution filters (.slnf) for building subsets. DO NOT USE FOR: single-project builds, incremental build issues (use incremental-build), compilation slowness within a project (use build-perf-diagnostics), non-MSBuild build systems. INVOKES: binlog MCP server tools (expensive_projects, expensive_targets, project_target_times); falls back to dotnet build -m, dotnet build /graph, binlog replay + grep.
The English summary is being prepared. Showing the original description.
Provides PyTorch-native distributed LLM pretraining using torchtitan with 4D parallelism (FSDP2, TP, PP, CP). Use when pretraining Llama 3.1, DeepSeek V3, or custom models at scale from 8 to 512+ GPUs with Float8, torch.compile, and distributed checkpointing.
The English summary is being prepared. Showing the original description.
Research a question by splitting it into two to five topics and assigning parallel documentation agents. Get a combined summary, detailed findings for each topic, and source URLs from web searches and documentation lookups.
Guide a development brief through specification, dependent tickets, implementation, review, and verification. Coordinate parallel work while keeping human approval at key decisions and recording agreed vocabulary, business rules, and the reasons behind technical choices.
Route code changes and SQL analysis to isolated Kortix work sessions with clear requirements and workspace details. Coordinate parallel investigations or independent pull request reviews, then summarize deliverables, verification results, and differences between findings.
Guides you through setting up RWKV, generating text, processing long documents in chunks, and fine-tuning models. Explains how to carry context between steps and compare memory use with Transformer models.
Helps decide whether to handle a task directly, create a subagent, or route it to an agent already familiar with the work. The choice depends on required background knowledge and whether plausible but incorrect results can be checked.
Coordinate temporary agents for code exploration, independent implementation tasks, verification, reviews, or complex debugging. Define self-contained assignments, run independent work in parallel, and check returned evidence before combining the results.
Convert .NET tests from xUnit.net v2 or v3 to MSTest v4 while keeping the existing test runner and target framework. Preserve assertions, shared setup, and parallel execution, then compare test counts and outcomes.
Operational guide for enabling hierarchical context parallelism in Megatron-Bridge, including config knobs, code anchors, pitfalls, and verification.
The English summary is being prepared. Showing the original description.