311 skills · sorted by relevance
Guide changes to SGLang runtime configuration and shared process state. Identify where resolved settings, runtime flags, resources, and per-forward state belong, and use the documented access patterns and checks when developing or testing.
Design file-based storage for agent context, including large tool results, plans, logs, and handoff notes. Keep information available outside the conversation and retrieve relevant sections when needed, instead of loading everything at once.
Plan and improve AI agent systems by managing the information models receive. This collection covers memory, agent coordination, tool design, context compression, evaluation, and the operating controls needed for reliable autonomous workflows.
Design structured summaries for long agent sessions that preserve goals, file changes, decisions, and next actions. Choose compression methods and timing, then test whether the agent can continue accurately without repeatedly retrieving information it already processed.
Preserve essential information during long agent tasks by extracting URLs, IDs, relevant fields, and conclusions into working notes. It also specifies where to record task handoff details before older tool results leave the agent’s context.
Check the `.planning/` directory for missing files, invalid settings, inconsistent state, and orphaned plans. Report issues to address, optionally request repairs, or check context usage—the share of the session’s text capacity already used.
Help a long-running agent reconnect with the intended product outcome after context compaction or a major work transition. Trace the current work hierarchy and read relevant source files to rebuild grounding before the next consequential action.
Long-context MoE training guidance for Megatron Bridge. Covers CP sizing, selective recompute, dispatcher choices, and practical patterns from DSV3, Qwen3, and Qwen3-Next long-context experiments.
The English summary is being prepared. Showing the original description.
Master optimization protocol for Gemini Agent (Antigravity) to unlock native 2M+ to 10M+ long-context reasoning, Gemini 4 Pro & Gemini 4 Flash dynamic thinking budget control, native context caching, Multimodal Live API, and high-speed problem solving / Protokol optimasi utama untuk Gemini Agent (Antigravity) untuk mengaktifkan pemikiran long-context 2M+ hingga 10M+, kontrol thinking budget dinamis Gemini 4 Pro & Gemini 4 Flash, context caching native, Multimodal Live API, dan pemecahan masalah kecepatan tinggi.
The English summary is being prepared. Showing the original description.
Elite AI context engineering specialist mastering dynamic context management, vector databases, knowledge graphs, and intelligent memory systems. Orchestrates context across multi-agent workflows, enterprise AI systems, and long-running projects with 2024/2025 best practices. Use PROACTIVELY for complex AI orchestration. Use when: the task directly matches context manager responsibilities within plugin agent-orchestration. Do not use when: a more specific framework or task-focused skill is clearly a better match.
The English summary is being prepared. Showing the original description.
Monitor, cap, and recover from context accumulation in agentic systems. Covers per-cycle cost tracking, context window auditing, budget caps with enforcement policies, emergency pruning when approaching limits, and progressive disclosure integration to minimize token spend on routing. Use when running long-lived agent loops (heartbeats, polling, autonomous workflows), when context windows are growing unpredictably between cycles, when API costs spike beyond expected baselines, when designing new agentic workflows that need cost guardrails from the start, or when post-mortem analysis reveals a cost incident caused by context accumulation.
The English summary is being prepared. Showing the original description.
Produces structured markdown session continuation documents for multi-session workflows. Captures active work streams, decisions with rationale, knowledge-file deltas learned during the session, the exact files to load next conversation, ingested content, artifacts, and open items into an ingestion-optimized handoff. Carries pending KF updates and unresolved items forward across handoff chains so context never drops. Generates a closeout checklist and echoes the starter prompt to chat. Use when approaching context limits, wrapping up a session, or managing work across conversations. Trigger on: "create a handoff," "session closeout," "wrap up this session," "build session handoff," "BSH," "continue next session," "we're running out of context," "context is getting long." Do NOT use for Memory optimization, context budget analysis, project audits, deciding whether work is ready for autonomous/Claude Code execution (use rootnode-handoff-trigger-check), or summaries with no continuation intent.
The English summary is being prepared. Showing the original description.
Find live webcams within a chosen map area around a port, airport, border crossing, city, or route. Retrieve a selected camera's thumbnail and player links, along with its update time, for visual context near a location.
Study Model Context Protocol (MCP), which connects AI applications with external tools, through guided lessons and exercises. Inspect requests and responses, practice implementation and security checks, and record progress alongside observed results.
Use Hindsight to recall past context, save reusable facts, and reason across accumulated memories. Supports continuity between conversations by preserving preferences, project decisions, and lessons learned, while helping explain the reasoning behind earlier choices.
Retrieve relevant Hindsight memories before planning or answering a request. Searches the agent’s own and shared memory stores, plus other relevant stores when needed, to bring past decisions, user preferences, and project context into the task.
Use OpenViking through OpenClaw to manage long-term memories, search archived conversations, and find imported resources or Agent Skills. Inspect recall traces to understand search results and recover full tool outputs stored outside the conversation.
Integrate Flash Attention into transformer models to reduce processing time and GPU memory use for long sequences. Covers PyTorch integration, advanced attention options, and benchmarks that compare speed, memory use, and output accuracy.
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.
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.
Optimize transformer attention, the calculation that relates tokens to one another, using Flash Attention. The skill guides PyTorch integration, GPU memory reduction, performance measurement, and output comparisons for training or inference with long sequences.
Creates comprehensive handoff documents for seamless AI agent session transfers. Triggered when: (1) user requests handoff/memory/context save, (2) context window approaches capacity, (3) major task milestone completed, (4) work session ending, (5) user says 'save state', 'create handoff', 'I need to pause', 'context is getting full', (6) resuming work with 'load handoff', 'resume from', 'continue where we left off'. Proactively suggests handoffs after substantial work (multiple file edits, complex debugging, architecture decisions). Solves long-running agent context exhaustion by enabling fresh agents to continue with zero ambiguity.
The English summary is being prepared. Showing the original description.
Elite AI context engineering specialist mastering dynamic context management, vector databases, knowledge graphs, and intelligent memory systems. Orchestrates context across multi-agent workflows, enterprise AI systems, and long-running projects with 2024/2025 best practices. Use PROACTIVELY for complex AI orchestration.
The English summary is being prepared. Showing the original description.
Decision guide for delegating to caveman-style subagents. Tells the main thread WHEN to spawn `cavecrew-investigator` (locate code), `cavecrew-builder` (1-2 file edit), or `cavecrew-reviewer` (diff review) instead of doing the work inline or using vanilla `Explore`. Subagent output is caveman-compressed so the tool-result injected back into main context is ~60% smaller — main context lasts longer across long sessions. Trigger: "delegate to subagent", "use cavecrew", "spawn investigator/builder/reviewer", "save context", "compressed agent output".
The English summary is being prepared. Showing the original description.