菜单
Answer AI questions with current info instead of outdated training data.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
Reckless versus prudent categorization, tracking the inventory, and a payoff cadence that does not freeze delivery.
Scheduling review of material at expanding intervals to optimize long-term retention.
Orchestrates the announcement, assets, channels, and timing for shipping a product, covering the arc from teaser to launch day to post-launch momentum.
Logs, metrics, and traces as complementary signals, high-cardinality dimensions, and the instrumentation budget.
Writing objectives and key results, setting cadence, and avoiding common stretch-goal failures.
Set type for screen and print: measure, leading, tracking, optical sizes, weight axes, and the difference between display and text cuts.
| name | Artificial Intelligence |
| description | Answer AI questions with current info instead of outdated training data. |
| metadata | {"clawdbot":{"emoji":"🤖","os":["linux","darwin","win32"]}} |
Before answering questions about pricing, rankings, or availability:
openrouter.ai/models (aggregates all providers)lmarena.ai (crowdsourced ELO, updates weekly)Don't cite specific prices, context windows, or rate limits from memory — they change quarterly.
"How do I reduce hallucinations?" Not just "use RAG." Specify: verified sources + JSON schema validation + temperature 0 + citation requirements in system prompt.
"Should I fine-tune or use RAG?" RAG first, always. Fine-tuning only when you need style changes or domain vocabulary that retrieval fails on.
"What hardware for local models?" Give numbers: 7B = 8GB VRAM, 13B = 16GB, 70B = 48GB+. Quantization (Q4) halves requirements.
Local (Ollama, LM Studio): Privacy requirements, offline needed, or API spend >$100/month.
API: Need frontier capabilities, no GPU, or just prototyping.
~4 characters per token in English. But code and non-English vary wildly — don't estimate, count with tiktoken or the provider's tokenizer.