This skill should be used when the user asks about Claude model strings, model IDs, model names, "which model", API calls to the Claude API, Anthropic SDK usage, Bedrock integration, Vertex AI integration, Azure AI Foundry, effort parameter, extended thinking, thinking budget, "migrate to opus", model migration, prompt adjustment for different models, tool overtriggering, prompt engineering for Claude, or when writing code that imports `anthropic` or `@anthropic-ai/sdk`.
From claude-craftnpx claudepluginhub metcalfc/claude-plugin --plugin claude-craftThis skill uses the workspace's default tool permissions.
references/effort-and-thinking.mdreferences/model-strings.mdreferences/prompt-patterns.mdDesigns and optimizes AI agent action spaces, tool definitions, observation formats, error recovery, and context for higher task completion rates.
Enables AI agents to execute x402 payments with per-task budgets, spending controls, and non-custodial wallets via MCP tools. Use when agents pay for APIs, services, or other agents.
Compares coding agents like Claude Code and Aider on custom YAML-defined codebase tasks using git worktrees, measuring pass rate, cost, time, and consistency.
Reference for Claude model strings, API configuration, and prompt patterns across all platforms. Use this when writing code that calls the Claude API, selecting models, or adjusting prompts for model behavioral differences.
The current generation is Claude 4.5/4.6. When users ask for "the latest model" or "best model", use these:
| Model | Model ID | Notes |
|---|---|---|
| Opus 4.6 | claude-opus-4-6-20250527 | Most capable, highest quality |
| Sonnet 4.6 | claude-sonnet-4-6-20250514 | Best balance of speed and quality |
| Haiku 4.5 | claude-haiku-4-5-20251001 | Fastest, lowest cost |
Shorthand aliases accepted by most SDKs: claude-opus-4-6, claude-sonnet-4-6, claude-haiku-4-5.
Model IDs differ by platform. The Anthropic 1P strings above work with the Anthropic SDK directly. For cloud providers:
us.anthropic.claude-opus-4-6-20250527-v1:0)claude-opus-4-6@20250527)See references/model-strings.md for the complete cross-platform table including previous-generation models.
Control response quality vs speed with the effort parameter (beta feature):
# Python SDK
client.messages.create(
model="claude-sonnet-4-6-20250514",
max_tokens=16000,
thinking={"type": "enabled", "budget_tokens": 10000},
betas=["effort-2025-11-24"],
output_config={"effort": "high"}, # low, medium, high
messages=[...]
)
| Effort | When to use |
|---|---|
low | Simple lookups, classification, formatting |
medium | General-purpose tasks (default behavior) |
high | Complex reasoning, math, code generation, analysis |
Effort interacts with thinking: high effort + thinking enabled = maximum reasoning depth. low effort may reduce or skip thinking even when enabled.
See references/effort-and-thinking.md for SDK examples in Python, TypeScript, and raw API, plus budget tuning guidance.
Different Claude models have different sensitivities. When migrating between models or tuning prompts:
Opus-class models are more eager to use tools. If tools are being called when they shouldn't be:
tool_choice: "auto" (not "any") and describe when NOT to use toolsClaude interprets "think" literally when extended thinking is enabled. The word "think" in prompts can trigger extended thinking unexpectedly:
Opus-class models tend toward comprehensive solutions. When you want focused, minimal changes:
Focus on the specific task. Do not refactor surrounding code, add features
beyond what was requested, or introduce abstractions for hypothetical future needs.
The right amount of complexity is the minimum needed for the current task.
<instructions>, <context>, <examples>) for structure — Claude responds well to themSee references/prompt-patterns.md for copy-paste snippets and detailed guidance.