Last Updated: 2025-10-24 19:21:55 EDT (UTC-04:00)
Agent Count: 65 agent definitions (verified)
Location: .claude/agents/
This directory serves as a reference point for the agent definitions used in the standards repository. The actual agent definitions are located in .claude/agents/ and follow Anthropic's agent delegation patterns.
Verified Count: 2025-10-24 19:21:55 EDT (UTC-04:00)
- Total Agent Definitions: 65 markdown files
- Categories: 18 directories
- Primary Location:
/home/william/git/standards/.claude/agents/
.claude/agents/
├── README.md # Agent directory index
├── analysis/ # Code analysis agents (3)
├── architecture/ # Architecture design agents (1)
├── consensus/ # Consensus coordination agents (7)
├── core/ # Core development agents (5)
├── data/ # Data processing agents (1)
├── development/ # Development workflow agents (1)
├── devops/ # DevOps & deployment agents (1)
├── documentation/ # Documentation agents (1)
├── github/ # GitHub integration agents (13)
├── hive-mind/ # Hive mind collective agents (1)
├── optimization/ # Performance optimization agents (5)
├── sparc/ # SPARC methodology agents (4)
├── specialized/ # Specialized domain agents (1)
├── swarm/ # Swarm coordination agents (3)
└── base-template-generator.md # Base template agent
Located in .claude/agents/core/
coder: Primary development implementationplanner: Project planning and task decompositionresearcher: Information gathering and analysisreviewer: Code review and quality assurancetester: Test creation and validation
Located in .claude/agents/github/
code-review-swarm: Automated PR reviewsgithub-modes: Multi-mode GitHub operationsissue-tracker: Issue management and triagemulti-repo-swarm: Cross-repository coordinationpr-manager: Pull request lifecycle managementproject-board-sync: Project board automationrelease-manager: Release coordinationrelease-swarm: Multi-repository releasesrepo-architect: Repository structure designswarm-issue: Swarm-based issue resolutionswarm-pr: Swarm-based PR managementsync-coordinator: Repository synchronizationworkflow-automation: CI/CD workflow automation
Located in .claude/agents/swarm/
adaptive-coordinator: Dynamic topology adaptationhierarchical-coordinator: Hierarchical task delegationmesh-coordinator: Peer-to-peer coordination
Located in .claude/agents/consensus/
byzantine-coordinator: Byzantine fault tolerancecrdt-synchronizer: Conflict-free replicated data typesgossip-coordinator: Gossip protocol coordinationperformance-benchmarker: Performance testingquorum-manager: Quorum-based decision makingraft-manager: Raft consensus protocolsecurity-manager: Security policy enforcement
Located in .claude/agents/optimization/
benchmark-suite: Comprehensive benchmarkingload-balancer: Task load distributionperformance-monitor: Real-time performance trackingresource-allocator: Resource allocation optimizationtopology-optimizer: Swarm topology optimization
Located in .claude/agents/sparc/
architecture: System architecture designpseudocode: Pseudocode specificationrefinement: Code refinement and optimizationspecification: Requirements specification
Located in .claude/agents/analysis/
code-analyzer: Static code analysiscode-review/analyze-code-quality: Quality metrics analysis- Additional analysis subdirectories
Agents are referenced in CLAUDE.md and used by Claude Code for task delegation:
## Agent Types for Task Tool (65 Available)
**Note**: These are conceptual agent types used with the Task tool,
not separate callable tools. Agent definitions are located in
`.claude/agents/` directory.Claude Code automatically delegates to agents based on:
- Keywords in user request: "test", "deploy", "document", "review"
- File patterns:
*.test.js→ testing agent,*.tf→ infrastructure agent - Task complexity: Multi-step tasks spawn coordinator agents
- Domain detection: Database queries → data agent, API endpoints → backend agent
Each agent type has specific tool access as defined in their respective markdown files:
- Development agents: Full file system access, code execution
- Testing agents: Test runners, coverage tools, limited write access
- Architecture agents: Read-only access, diagram generation
- Documentation agents: Markdown tools, read access, limited write to docs/
- DevOps agents: Infrastructure tools, deployment scripts, environment access
- Analysis agents: Read-only access, static analysis tools
Agents complement the skills system:
- Skills (61 total): Domain knowledge and implementation patterns
- Agents (65 total): Task delegation and workflow orchestration
Skills provide the "what" (standards, patterns, best practices) Agents provide the "how" (task execution, coordination, tooling)
All agent definitions follow these standards:
- Standard: NIST ET (America/New_York)
- Format: ISO 8601 with timezone offset
- Example:
2025-10-24 19:21:55 EDT (UTC-04:00)
- Version number
- Last updated timestamp (NIST ET)
- Description and purpose
- Tool restrictions
- Delegation triggers
Last Audit: 2025-10-24 19:21:55 EDT (UTC-04:00)
# Verify agent count
find .claude/agents -type f -name "*.md" | wc -l
# Output: 65
# List agent categories
ls -1 .claude/agents/
# Output: 18 directories + filesThe main CLAUDE.md file references these agents and provides:
- Agent type categorization
- Usage examples
- Coordination protocols
- MCP tool integration
- Concurrent execution patterns
When migrating or updating agents:
- Preserve datetime format: Always use NIST ET timezone
- Update counts: Run verification commands after changes
- Document changes: Update README with timestamp
- Validate structure: Ensure agent definitions follow standards
- Test delegation: Verify automatic delegation triggers work
- Main Router: CLAUDE.md
- Agent Index: .claude/agents/README.md
- Skills System: skills/README.md
# Count agents
find .claude/agents -type f -name "*.md" | wc -l
# List agent categories
ls -1 .claude/agents/
# Check for datetime format compliance
grep -r "EDT\|EST" .claude/agents/ | head -5Maintained by: Standards Repository Team Repository: https://github.com/williamzujkowski/standards Datetime Standard: NIST ET (America/New_York) - ISO 8601 format