principle-guard-the-context-window
Apply when context is filling up: large outputs, long files, repeated reads, fan-out planning. Route bulk to subagents; keep summaries in the main thread, not…
Skills
Find an Agent Skill for your task. Read the guide and copy the install command.
Updated 11-11 00:16
Agents & models
26,089 results
principle-never-block-on-the-human
Apply when tempted to ask 'should I do X?' on reversible work. Proceed, present the result, let the human course-correct after the fact; reserve confirmation…
swarm
Fan out N parallel workers, drain them, and return one report. Use for /swarm, 'swarm this', or parallel coverage, races, gauntlets, and exploration.
ai-engineer
Expert AI Engineer role (10+ Years Exp). Focuses on production-grade GenAI, Agentic Systems, Advanced RAG, and rigorous Evaluation.
skill-creator
Guide for creating Agent Skills that extend Antigravity's capabilities. Use when creating new skills, updating existing skills, or packaging skills for…
sequential-thinking
Use when complex problems require systematic step-by-step reasoning with ability to revise thoughts, branch into alternative approaches, or dynamically adjust…
specs
Business-rule specs for KiloClaw billing/lifecycle/controller/data model/Composio, MCP Gateway auth, model experiments, Security Agent, subscription center,…
configuring-dbt-mcp-server
Generates MCP server configuration JSON, resolves authentication setup, and validates server connectivity for dbt. Use when setting up, configuring, or…
memory-3layer
Establish platform-neutral three-layer memory for long-horizon, cross-session Agent work. Use it to load project memory, record reusable facts and daily progress, maintain tacit knowledge, migrate legacy .claude/memory, check memory status, or clean up outdated entries; Claude Code and Codex can be wired automatically through their respective Hooks, while other Agent Skills…
meta-skill-creator
Create, refactor, or accept reusable skill packages. Suitable for turning repeated workflows into cross-host capability packages, and clarifying trigger rules, first actions, progressive loading, asset templates, script validation, trigger evaluation, baseline comparison, acceptance evidence, closed-loop governance, public delivery boundaries, and unforgeable proof of execution; also suitable for cleaning up redundant references, separating human-readable templates from machine-structured data, filling in acceptance records before the first public release, recording runtime feedback, and generating write-back or no-write-back decisions. When the user mentions "...
meta-skill-creator
Create, refactor, or accept reusable skill packages. It applies to turning repetitive workflows into cross-host capability packages, and to clarifying trigger rules, first actions, progressive loading, asset templates, script validation, trigger evaluation, baseline comparison, acceptance evidence, closed-loop governance, public delivery boundaries, and unforgeable proof of execution; it also applies to cleaning up redundant references, splitting human-readable templates from machine-structured data, filling in acceptance records before first public release, recording run feedback, and generating write-back or no-write-back decisions. When the user mentions "…
deepspeed
Expert guidance for distributed training with DeepSpeed - ZeRO optimization stages, pipeline parallelism, FP16/BF16/FP8, 1-bit Adam, sparse attention
dspy
Build complex AI systems with declarative programming, optimize prompts automatically, create modular RAG systems and agents with DSPy - Stanford NLP's…
implementing-llms-litgpt
Implements and trains LLMs using Lightning AI's LitGPT with 20+ pretrained architectures (Llama, Gemma, Phi, Qwen, Mistral). Use when need clean model…
query
Query Lode's decision replay index for why a project chose a path, what alternatives were rejected, what should be revisited, what impact a decision had, or…
agent-autoloop
Configure the Agent automation hub: add rules, connect new event plugins, connect new Agents, connect new channels (Feishu, QQ, DingTalk, WeCom...), write handling manuals, deploy to servers, multiple computers. When the user says "let the Agent handle things itself when something happens" "have the Agent do something at a certain time every day" "connect GitHub / alerts /...
zhihu-search
Zhihu Open Platform search capability integration — use when an Agent needs to search Zhihu content, web-wide information, direct answers, or view the hot list. Includes daily quota management and intelligent source selection strategy.
996-workhorse
Triggered when the user expresses a desire for the AI to work as intensely as 996, be on standby around the clock, and keep fighting continuously without getting tired. This includes but is not limited to: 'Can you work a little harder', 'work like 996', 'like 007', 'work me to death', 'don't stop', 'continue, don't stop', 'keep going', 'don't stop', 'work harder', 'more…
harness-build-core-manifest
Build a core rules checklist of ≤60 lines to prevent Agent behavior from drifting at the source
find-skills-sh
Helps users discover and install agent skills when they ask questions like "how do I do X", "find a skill for X", "is there a skill that can...", or express…
find-skills
Helps users discover and evaluate reusable skills when they ask how to do X, want a skill for X, or want to extend Codex with a reusable workflow.
handoff
Use when the user wants to transfer the current conversation, task state, project context, decisions, constraints, or next steps to another agent or a new…
yomiyasu
An Agent Skill that rewrites unnatural Japanese generated by AI into natural sentences that are easy for humans to read and information-dense. Use it for requests such as "make this text easier to read," "remove the AI-ness," "eliminate the AI smell," "make it natural Japanese," and "deodorize the text," as well as when polishing technical articles, business specifications/PR descriptions, and essays/note posts. Decomposing inanimate subjects...
acceptance-orchestrator
Use when a coding task should be driven end-to-end from issue intake through implementation, review, deployment, and acceptance verification with minimal human…