其他技能

archive ステップ完了後に Lessons を分析し constitution.md への追加提案を生成するサブエージェント

中文简介正在准备中。当前显示原文说明。

Guide for using renderers — the bridge between chat-style messages and token sequences. Covers renderer setup, TrainOnWhat, vision inputs, model family matching, and custom renderers. Use when the user asks about renderers, tokenization, message formatting, or vision inputs.

中文简介正在准备中。当前显示原文说明。

Guide for checkpointing — saving, loading, and resuming training with CheckpointRecord. Use when the user asks about saving weights, resuming training, checkpoint management, or the checkpoint lifecycle.

中文简介正在准备中。当前显示原文说明。

Guide for using completers — TokenCompleter and MessageCompleter for text generation during RL rollouts and evaluation. Use when the user asks about generating text, completing messages, or using completers in RL environments.

中文简介正在准备中。当前显示原文说明。

Guide for evaluation — inline evaluators, Inspect AI integration, and custom evaluators for measuring training progress. Use when the user asks about evaluation, metrics, benchmarks, or how to measure model quality during training.

中文简介正在准备中。当前显示原文说明。

Guide for the weight lifecycle — downloading trained weights from Tinker, merging LoRA adapters into HuggingFace models, and publishing to HuggingFace Hub. Use when the user asks about exporting, downloading, merging, or publishing trained model weights.

中文简介正在准备中。当前显示原文说明。

Improve interface resilience through better error handling, i18n support, text overflow handling, and edge case management, making interfaces robust and production-ready. Use when the user asks to harden, make production-ready, handle edge cases, add error states, or fix overflow and i18n issues.

中文简介正在准备中。当前显示原文说明。

Clarify priority and emphasis by fixing unclear primary actions, competing visual weight, loud section titles, and weak label/value treatment. Use when the user mentions hierarchy, prioritization, too much competition, grayscale confusion, loud headings, or unclear primary actions.

中文简介正在准备中。当前显示原文说明。

Perform a final quality pass fixing alignment, spacing, consistency, and micro-detail issues before shipping. Use when the work is functionally complete and needs finishing touches—not when the hierarchy, structure, tone, or technical foundation still need major changes.

中文简介正在准备中。当前显示原文说明。

Add or refine depth through elevation systems, raised and inset surfaces, layered overlap, and meaningful shadow behavior so interfaces feel structured instead of flat. Use when the user mentions depth, elevation, shadows, layering, flat cards, pressed states, drag feedback, or inset controls.

中文简介正在准备中。当前显示原文说明。

Improve layout composition, spacing systems, grouping, and visual rhythm. Use when the user mentions weak layout structure, arbitrary spacing, weak grouping, crowded composition, or layout monotony rather than priority/emphasis problems.

中文简介正在准备中。当前显示原文说明。

Add moments of joy, personality, and surprise through success states, empty states, loading moments, hover details, and celebratory polish once the fundamentals already work. Use when the user wants delight, charm, personality, or memorable micro-interactions—not when the main need is baseline cleanup or stronger visual impact.

中文简介正在准备中。当前显示原文说明。

Use when: Router bridges your team and the remote team across Telegram. Classifies inbound messages from the remote agent, dispatches Builder or Tester as subagents, composes replies, and posts outbound state-change notifications.

中文简介正在准备中。当前显示原文说明。

Context-survival and proactive-iteration extensions. Always-on. Covers Working Buffer Protocol, Compaction Recovery, Relentless Resourcefulness, and Self-Improvement Guardrails.

中文简介正在准备中。当前显示原文说明。

Use when: planner mode, planning mode, proposal, scope definition, estimate, timeline, how hard, feasibility, effort, how long, how big, client message, external communication, pricing, delivery confidence, commercial risk, or project delivery planning.

中文简介正在准备中。当前显示原文说明。

Use when: Good night Team, EOD, Wrapping up, or end-of-day consolidation. Archives session context, promotes durable patterns, and compacts live agent memory files for the next session.

中文简介正在准备中。当前显示原文说明。

Use when: Good morning Team, start-of-day briefing, or morning status request. Architect orchestrates; Builder, Tester, and Router return their sections as subagents.

中文简介正在准备中。当前显示原文说明。

Shared foundation for all human-voice writing skills. Load before any profession-specific skill. Contains: burstiness principle, perplexity principle, soul injection techniques, community-discovered AI patterns with sources, additional core patterns, banned AI vocabulary, and scoring rubric.

中文简介正在准备中。当前显示原文说明。

Convergence loop that iteratively improves work until quality thresholds are met. Scores, diagnoses specific weaknesses, rewrites, re-scores, and repeats until all dimensions reach 8/10 or diminishing returns are detected. Use when you want to refine or iterate on work, after the analyst gives a REVISE verdict, or when output quality is below your bar. Part of the architect-system loop. Outputs to system/refinery-log.md.

中文简介正在准备中。当前显示原文说明。

Convergence loop that iteratively improves work until quality thresholds are met. Use when: (1) "refine this", "make this better", "iterate on this", (2) After /analyst gives a REVISE verdict, (3) Output quality is below 8/10, (4) "run the refinery loop on this". Reads analyst reviews from system/reviews/. Generates, scores, diagnoses, rewrites, re-scores. Stops when all dimensions >= 8/10 or diminishing returns detected. Outputs to system/refinery-log.md.

中文简介正在准备中。当前显示原文说明。

Track build processing, find latest builds, and clean up old builds with asc. Use when managing build retention or waiting on processing.

中文简介正在准备中。当前显示原文说明。

Triage TestFlight crashes, beta feedback, and performance diagnostics using asc. Use when the user asks about TF crashes, TestFlight crash reports, beta tester feedback, app hangs, disk writes, launch diagnostics, or wants a crash summary for a build or app.

中文简介正在准备中。当前显示原文说明。

Capture and interpret iOS Simulator ETTrace profiles. Use when profiling launch or runtime latency, comparing traces, or finding CPU-heavy stacks.

中文简介正在准备中。当前显示原文说明。

This skill should be used when the user asks to "analyze network traffic with Wireshark", "capture packets for troubleshooting", "filter PCAP files", "follow TCP/UDP streams", "detect network anomalies", "investigate suspicious traffic", or "perform protocol analysis". It provides comprehensive techniques for network packet capture, filtering, and analysis using Wireshark.

中文简介正在准备中。当前显示原文说明。