Skip to main content

Anthropic Agents SDK Reference Bibliography

Standards Development Reference for Skills, Commands, Hooks, Plugins, Subagents, MCP, and Memory​

Version 1.0 - December 2025

Table of Contents​

  1. Agent SDK Overview
  2. Agent Skills
  3. Slash Commands
  4. Hooks
  5. Plugins & Marketplaces
  6. Subagents
  7. Model Context Protocol (MCP)
  8. Memory & CLAUDE.md
  9. Settings & Configuration
  10. SDK Implementation (TypeScript/Python)
  11. Best Practices & Engineering Guides
  12. Community Resources

1. Agent SDK Overview​

Core documentation for understanding the Claude Agent SDK architecture and capabilities.

ResourceURLDescriptionPriority
Agent SDK Overview (Official)https://docs.claude.com/en/api/agent-sdk/overviewPrimary SDK documentation with features, authentication, and configurationCRITICAL
Building Agents with the Claude Agent SDKhttps://www.anthropic.com/engineering/building-agents-with-the-claude-agent-sdkEngineering blog explaining SDK design principles and agent buildingCRITICAL
Claude Agent SDK (npm)https://www.npmjs.com/package/@anthropic-ai/claude-agent-sdkTypeScript SDK package with installation instructionsHIGH
Claude Agent SDK Python (GitHub)https://github.com/anthropics/claude-agent-sdk-pythonPython SDK repository with examples and API referenceHIGH
Claude Agent SDK TypeScript (GitHub)https://github.com/anthropics/claude-agent-sdk-typescriptTypeScript SDK repository with examplesHIGH
Claude Code Main Repositoryhttps://github.com/anthropics/claude-codeMain Claude Code repository containing plugins and documentationHIGH
Platform SDK Documentationhttps://platform.claude.com/docs/en/agent-sdk/overviewAlternative documentation portal for SDKMEDIUM

2. Agent Skills​

Documentation for creating, managing, and deploying Agent Skills (SKILL.md).

ResourceURLDescriptionPriority
Agent Skills Overviewhttps://docs.claude.com/en/docs/agents-and-tools/agent-skills/overviewOfficial skills overview with structure, invocation, and securityCRITICAL
Agent Skills in Claude Codehttps://docs.claude.com/en/docs/claude-code/skillsClaude Code-specific skills documentation with examplesCRITICAL
Equipping Agents with Agent Skillshttps://www.anthropic.com/engineering/equipping-agents-for-the-real-world-with-agent-skillsEngineering deep-dive on skill architecture and progressive disclosureCRITICAL
Introducing Agent Skills (Blog)https://www.anthropic.com/news/skillsOfficial announcement with feature overviewHIGH
Skills Repository (GitHub)https://github.com/anthropics/skillsOfficial skills repository with examples and templatesHIGH
Skills API Quickstarthttps://docs.claude.com/en/docs/agents-and-tools/agent-skills/quickstartAPI integration guide for skillsHIGH

Skills File Structure Reference​

ComponentLocationDescription
SKILL.md{skill-dir}/SKILL.mdMain skill definition with YAML frontmatter
Reference files{skill-dir}/*.mdAdditional documentation loaded on-demand
Scripts{skill-dir}/scripts/Executable scripts for tool capabilities
Templates{skill-dir}/templates/Template files for skill outputs

Skills YAML Frontmatter Fields​

FieldRequiredDescription
nameYesSkill identifier
descriptionYesTrigger description for Claude's context
allowed-toolsNoTools available when skill is active
when_to_useNo(Undocumented) Activation guidance

3. Slash Commands​

Documentation for custom slash commands in .claude/commands/.

ResourceURLDescriptionPriority
Slash Commands Referencehttps://docs.anthropic.com/en/docs/claude-code/slash-commandsComplete slash commands documentationCRITICAL
Built-in Commandshttps://docs.claude.com/en/docs/claude-code/cli-referenceCLI reference including built-in slash commandsHIGH
Custom Commands Guidehttps://aiengineerguide.com/blog/claude-code-custom-command/Tutorial on creating custom commandsMEDIUM

Slash Commands Directory Structure​

ScopeLocationDescription
Project.claude/commands/*.mdProject-specific commands (shows "(project)")
User~/.claude/commands/*.mdPersonal commands across all projects (shows "(user)")
Namespaced.claude/commands/{namespace}/*.mdOrganized commands with /namespace:command syntax

Command Markdown Features​

FeatureSyntaxDescription
Arguments$ARGUMENTSPlaceholder for command arguments
File Reference@path/to/fileInclude file contents in command
Bash Execution!commandExecute bash and include output
Frontmatter---\nallowed-tools: [...]\ndescription: ...\n---Command metadata

4. Hooks​

Documentation for lifecycle hooks in .claude/settings.json.

ResourceURLDescriptionPriority
Hooks Referencehttps://docs.claude.com/en/docs/claude-code/hooksComplete hooks technical referenceCRITICAL
Hooks Quickstarthttps://docs.claude.com/en/docs/claude-code/hooks-guideGetting started with hooksCRITICAL
Hooks Mastery Repositoryhttps://github.com/disler/claude-code-hooks-masteryCommunity hook examples and patternsMEDIUM

Hook Events Reference​

EventTriggerCan BlockUse Cases
PreToolUseBefore tool executionYes (exit 2)Validation, security, logging
PostToolUseAfter tool completionNoFormatting, cleanup, notifications
UserPromptSubmitUser submits promptYesPrompt validation, context injection
NotificationClaude needs attentionNoCustom notifications
StopClaude finishes respondingYes ("block")Completion messages, TTS
SubagentStopSubagent completesYesSubagent coordination
PreCompactBefore context compactionNoBackup, logging
SessionStartSession begins/resumesNoContext loading, setup
SessionEndSession endsNoCleanup, persistence

Hook Configuration Schema​

{
  "hooks": {
    "EventName": [
      {
        "matcher": "ToolPattern|Regex",
        "hooks": [
          {
            "type": "command|prompt",
            "command": "shell-command",
            "prompt": "llm-evaluation-prompt"
          }
        ]
      }
    ]
  }
}

Hook Exit Codes​

Exit CodeMeaningBehavior
0SuccessContinue normally, stdout shown in transcript
1ErrorContinue, stderr logged
2BlockBlock tool/action, stderr sent to Claude

5. Plugins & Marketplaces​

Documentation for plugin system and marketplace distribution.

ResourceURLDescriptionPriority
Plugins Announcementhttps://www.anthropic.com/news/claude-code-pluginsOfficial plugin feature announcementHIGH
Plugins Referencehttps://anthropic.mintlify.app/en/docs/claude-code/plugins-referenceComplete plugin technical referenceCRITICAL
Plugin Marketplaceshttps://docs.claude.com/en/docs/claude-code/plugin-marketplacesMarketplace creation and distributionCRITICAL
Plugins in the SDKhttps://platform.claude.com/docs/en/agent-sdk/pluginsSDK plugin integrationHIGH
Official Plugins (GitHub)https://github.com/anthropics/claude-code/blob/main/plugins/README.mdAnthropic's example pluginsHIGH
Claude Code Plugins Marketplacehttps://claudecodemarketplace.com/Community plugin directoryMEDIUM
Claude Plugins CLIhttps://claude-plugins.dev/CLI for plugin installationMEDIUM

Plugin Directory Structure​

my-plugin/
├── .claude-plugin/
│   └── plugin.json          # Plugin manifest (required)
├── commands/                 # Slash commands
│   └── my-command.md
├── agents/                   # Subagent definitions
│   └── my-agent.md
├── skills/                   # Agent skills
│   └── my-skill/
│       └── SKILL.md
├── hooks/                    # Hook configurations
│   └── hooks.json
└── mcp-servers/             # MCP server configs
    └── servers.json

Plugin Manifest (plugin.json) Schema​

FieldRequiredDescription
nameYesPlugin identifier
versionYesSemver version
descriptionYesPlugin description
authorNoAuthor info (name, email, url)
commandsNoCustom command paths
agentsNoAgent directory or paths
skillsNoSkills directory
hooksNoHooks configuration file
mcpServersNoMCP configuration file

Marketplace Commands​

CommandDescription
/plugin marketplace add user/repoAdd a marketplace
/plugin install plugin@marketplaceInstall from marketplace
/plugin marketplace update nameRefresh marketplace
/pluginBrowse installed plugins

6. Subagents​

Documentation for subagent orchestration in .claude/agents/.

ResourceURLDescriptionPriority
Subagents Documentationhttps://docs.anthropic.com/en/docs/claude-code/sub-agentsOfficial subagents referenceCRITICAL
Subagents in the SDKhttps://docs.claude.com/en/docs/claude-code/sdk/subagentsSDK subagent integrationHIGH
Community Subagentshttps://subagents.cc/Community subagent collectionMEDIUM
Subagents Repositoryhttps://github.com/wshobson/agents85+ specialized agents collectionMEDIUM

Subagent Directory Structure​

ScopeLocationDescription
Project.claude/agents/*.mdProject-specific agents
User~/.claude/agents/*.mdPersonal agents across projects

Subagent Markdown Schema​

---
name: agent-name
description: When and why to use this agent
tools: Read, Write, Bash, Grep, Glob  # Optional, inherits if omitted
model: sonnet|opus                     # Optional model override
---

# Agent Name

System prompt and behavioral instructions...

## Capabilities
- Specific capability 1
- Specific capability 2

## Context
Additional context and examples...

Built-in Subagents​

AgentPurposeMode
PlanCodebase research for planningPlan mode only
ExploreFast read-only codebase explorationExecution mode

Subagent Management Commands​

CommandDescription
/agentsInteractive agent management
/agents createCreate new agent with guided prompts
/agents listList available agents

7. Model Context Protocol (MCP)​

Documentation for MCP server integration.

ResourceURLDescriptionPriority
MCP in Claude Codehttps://docs.anthropic.com/en/docs/claude-code/mcpClaude Code MCP integrationCRITICAL
MCP in the SDKhttps://docs.claude.com/en/docs/claude-code/sdk/sdk-mcpSDK MCP configurationHIGH
Introducing MCPhttps://www.anthropic.com/news/model-context-protocolOfficial MCP announcementHIGH
Desktop Extensionshttps://www.anthropic.com/engineering/desktop-extensionsOne-click MCP installationHIGH
MCP GitHub Organizationhttps://github.com/modelcontextprotocolOfficial MCP repositoriesHIGH
MCP Coursehttps://anthropic.skilljar.com/introduction-to-model-context-protocolAnthropic Academy MCP courseMEDIUM

MCP Configuration Locations​

ScopeLocationDescription
Project.mcp.jsonProject MCP servers (version controlled)
Local.claude/settings.local.jsonLocal MCP config (gitignored)
User~/.claude/settings.jsonUser-wide MCP servers
EnterpriseManaged policyOrganization MCP controls

MCP Server Configuration Schema​

{
  "mcpServers": {
    "server-name": {
      "command": "npx|node|python",
      "args": ["server-package-or-script"],
      "env": {
        "ENV_VAR": "value",
        "DYNAMIC": "${SHELL_VAR:-default}"
      }
    }
  }
}

MCP Server Types​

TypeConfigurationDescription
stdiocommand + argsExternal process via stdin/stdout
ssetype: "sse" + urlHTTP/SSE remote server
sdkIn-processSDK MCP servers (Python/TS)

MCP Tool Naming Convention​

PatternExampleDescription
mcp__server__toolmcp__github__create_issueServer-namespaced tool
mcp__list_resources-List available resources
mcp__read_resource-Read a specific resource

8. Memory & CLAUDE.md​

Documentation for persistent context and memory files.

ResourceURLDescriptionPriority
Memory Managementhttps://docs.claude.com/en/docs/claude-code/memoryOfficial CLAUDE.md documentationCRITICAL
Bringing Memory to Teamshttps://www.anthropic.com/news/memoryMemory feature announcementHIGH
Context Managementhttps://anthropic.com/news/context-managementContext editing and memory toolsHIGH

Memory File Locations​

FileLocationDescription
User Memory~/.claude/CLAUDE.mdPersonal preferences (all projects)
Project Memory./CLAUDE.md or ./.claude/CLAUDE.mdProject context (committed)
Local Memory./CLAUDE.local.mdPersonal project notes (gitignored)

Memory File Features​

FeatureSyntaxDescription
File Import@path/to/file.mdImport external content
Absolute Import@~/path/from/homeImport from home directory
Recursive DiscoveryAutomaticFinds CLAUDE.md up directory tree
Subtree DiscoveryAutomaticLoads nested CLAUDE.md on file access

Memory Best Practices​

PracticeDescription
Be specific"Use 2-space indentation" vs "Format code properly"
Use structureGroup related items under markdown headings
Keep leanInclude only essential session context
Use importsReference detailed docs instead of inline

9. Settings & Configuration​

Documentation for Claude Code settings files.

ResourceURLDescriptionPriority
Settings Documentationhttps://docs.claude.com/en/docs/claude-code/settingsOfficial settings referenceCRITICAL
Configuration Guidehttps://claudelog.com/configuration/Comprehensive configuration guideMEDIUM
CLI Cheatsheethttps://shipyard.build/blog/claude-code-cheat-sheet/Settings and CLI referenceMEDIUM

Settings File Hierarchy (Precedence Low→High)​

LevelLocationDescription
User~/.claude/settings.jsonGlobal user settings
Project.claude/settings.jsonProject settings (committed)
Local.claude/settings.local.jsonLocal overrides (gitignored)
EnterpriseManaged policyOrganization policies

Key Settings Fields​

FieldDescription
modelDefault model selection
maxTokensMaximum response tokens
permissions.allowedToolsPre-approved tools
permissions.denyDenied tool patterns
hooksHook configurations
mcpServersMCP server configs
extraKnownMarketplacesPlugin marketplaces
enabledPluginsAuto-installed plugins

10. SDK Implementation (TypeScript/Python)​

Detailed SDK implementation references.

ResourceURLDescriptionPriority
TypeScript SDK (npm)https://www.npmjs.com/package/@anthropic-ai/claude-agent-sdkTypeScript packageHIGH
Python SDK (PyPI)https://pypi.org/project/claude-agent-sdk/Python packageHIGH
SDK Overviewhttps://docs.anthropic.com/en/docs/claude-code/sdkSDK feature overviewHIGH
SDK Tutorial (DataCamp)https://www.datacamp.com/tutorial/how-to-use-claude-agent-sdkStep-by-step SDK tutorialMEDIUM
SDK Guide (Bind AI)https://blog.getbind.co/2025/10/03/how-to-create-agents-with-claude-agents-sdk/Comprehensive SDK guideMEDIUM

SDK Query Methods​

MethodLanguageDescription
query()BothSimple async generator for queries
ClaudeSDKClientBothFull client with tools, hooks, sessions

SDK Configuration Options​

OptionDescription
promptUser query
system_promptSystem instructions
allowedToolsTools to enable
disallowedToolsTools to block
permissionModePermission handling
mcpConfigMCP configuration path
mcpServersInline MCP servers
pluginsPlugin configurations
maxTurnsMaximum agent turns

11. Best Practices & Engineering Guides​

Curated guides for production implementation.

ResourceURLDescriptionPriority
Claude Code Best Practiceshttps://www.anthropic.com/engineering/claude-code-best-practicesOfficial best practicesCRITICAL
Complete Guide (Sid Bharath)https://www.siddharthbharath.com/claude-code-the-complete-guide/Comprehensive usage guideHIGH
Agent Skills Deep Divehttps://leehanchung.github.io/blogs/2025/10/26/claude-skills-deep-dive/Technical skills analysisMEDIUM
Claude Code Hooks (GitButler)https://blog.gitbutler.com/automate-your-ai-workflows-with-claude-code-hooksHooks automation guideMEDIUM
WorkflowDescription
TDD with ClaudeTests first, then implementation
Plan ModeResearch before execution
Subagent DelegationIsolate specialized tasks
Context ManagementRegular /clear and /compact

12. Community Resources​

Community-maintained resources and collections.

ResourceURLDescriptionPriority
Awesome Claude Codehttps://github.com/hesreallyhim/awesome-claude-codeCurated commands and workflowsMEDIUM
Claude Command Suitehttps://github.com/qdhenry/Claude-Command-Suite148+ slash commandsMEDIUM
WShobson Commandshttps://github.com/wshobson/commands57 production commandsMEDIUM
WShobson Agentshttps://github.com/wshobson/agents85 specialized agentsMEDIUM
Claude Code Plugins Plushttps://github.com/jeremylongshore/claude-code-plugins-plus243+ plugins marketplaceMEDIUM
My Claude Code Setuphttps://github.com/centminmod/my-claude-code-setupStarter template with hooksMEDIUM
ClaudeLoghttps://claudelog.com/Docs, guides, and best practicesMEDIUM

Standards Development Checklist​

Use this checklist when building standards for each component:

For Each Standard Document, Include:​

  • Purpose & Scope: What problem does this solve?
  • File Locations: Where files must be placed
  • Schema Definition: Required and optional fields
  • Naming Conventions: How to name files, commands, agents
  • Security Considerations: Permissions, sandboxing, trust
  • Examples: Minimal and comprehensive examples
  • Anti-Patterns: What to avoid
  • Testing Guidelines: How to validate
  • Version Compatibility: SDK version requirements
  • Migration Paths: Upgrading from previous versions

Document Metadata​

FieldValue
CreatedDecember 2025
SourcesOfficial Anthropic Documentation
Total References90+ URLs
Primary Domainsdocs.claude.com, anthropic.com, github.com/anthropics

This bibliography is intended for internal standards development. All URLs were verified at time of creation. For the most current information, always reference the official Anthropic documentation.