github/gh-aw

• **Archetype**: harness-framework
• **Stars**: 4,288
• **Last pushed**: 2026-04-13
• **License**: MIT
• **Discovered**: 2026-04-13
• **Source**: backlog-processing
• **Skills found**: 22 (console-rendering, custom-agents, developer, dictation, documentation, error-messages, error-pattern-safety, error-recovery-patterns, gh-agent-session, gh-agent-task, github-copilot-agent-tips-and-tricks, github-discussion-query, github-issue-query, github-mcp-server, github-pr-query, github-script, http-mcp-headers, javascript-refactoring, messages, reporting, skillz-integration, temporary-id-safe-output)

Summary

GitHub's official "Agentic Workflows" framework that enables writing agent workflows in natural language markdown and executing them in GitHub Actions. Built in Go with comprehensive CLI tooling, includes 22 skills covering GitHub automation, error recovery patterns, MCP server integration, agent sessions, and safety guardrails. Provides a complete harness for agentic GitHub automation with production-ready error handling and recovery strategies.

Assessment

VERY HIGH VALUE. This is GitHub's authoritative framework for agentic workflow orchestration with substantial procedural knowledge. The error recovery patterns are particularly valuable - containing circuit breaker implementations, exponential backoff strategies, and systematic debugging approaches. The GitHub automation skills encode official patterns for issue/PR queries, discussion management, and MCP server integration. The safety and guardrail patterns are directly applicable to babysitter's breakpoint and validation systems.

Extraction Priority

VERY HIGH - Contains official GitHub workflow orchestration patterns that are directly transferable:

• Error recovery and retry patterns -> specializations/shared/
• Agentic workflow orchestration -> methodologies/agentic-workflows/
• GitHub automation processes -> specializations/devops-sre-platform/
• Safety and guardrail patterns -> specializations/shared/

Skills Inventory

Processes

- Source: skills/error-recovery-patterns/SKILL.md (lines 25-50) - Placement: specializations/shared/ - Inputs: Operation type, failure modes, retry constraints - Outputs: Recovery strategy, retry configuration, circuit breaker setup - Complexity: complex - Notes: Reduces retry loops to <10%, includes fail-fast for non-transient errors

• **error-recovery-pipeline**: Comprehensive error handling with circuit breakers and exponential backoff

- Source: Overall framework design and documentation - Placement: methodologies/agentic-workflows/ - Inputs: Workflow requirements, safety constraints, execution environment - Outputs: Workflow definition, execution plan, safety guardrails - Complexity: complex

• **agentic-workflow-orchestration**: Process for designing and executing natural language agent workflows

- Source: github-issue-query, github-pr-query, github-discussion-query skills - Placement: specializations/devops-sre-platform/ - Inputs: Repository, automation requirements, API access - Outputs: Automation scripts, query patterns, integration configuration - Complexity: moderate

• **github-automation-pipeline**: Systematic approach to GitHub repository automation

Plugin Ideas

- What install.md would do: Install GitHub CLI, configure agentic workflows, set up error recovery patterns, create GitHub Actions templates - Processes it would copy: error-recovery-pipeline, github-automation-pipeline, agentic-workflow-orchestration - Configs/hooks it would create: GitHub Actions workflows, error recovery configs, safety guardrails, MCP server configurations - Source evidence: Complete framework with 22 skills for GitHub automation and agentic workflows

• **agentic-github-workflows**: Complete GitHub workflow automation setup

- What install.md would do: Install circuit breaker libraries, configure retry patterns, set up monitoring for failure rates - Processes it would copy: error-recovery-pipeline, error-classification-process - Configs/hooks it would create: Retry configurations, circuit breaker settings, error monitoring dashboards - Source evidence: Comprehensive error recovery skills with production metrics (target <10% retry loops)

• **error-recovery-suite**: Robust error handling and recovery infrastructure

Harness Integration Ideas

Harness Adapter for GitHub Agentic Workflows

**Capability Assessment for Babysitter Integration:**

**Integration Viability:** PARTIAL - Excellent tool ecosystem and MCP integration but **completely lacks stop hooks** needed for babysitter's orchestration interruption model.

• **Adapter implementation**: createGhAwAdapter in packages/sdk/src/harness/adapters/
• **Plugin structure**: plugins/babysitter-gh-aw/ for GitHub Agentic Workflows integration
• **CLI integration**: gh aw command patterns, workflow execution, agent session management
• **Major limitation**: **No interruption mechanisms** - would require custom workflow pause/resume implementation

TUI/Orchestration Improvement: Agentic Workflow Patterns

• **Current limitation**: Our harness lacks natural language workflow definition and GitHub Actions integration
• **Integration approach**: Adapt GitHub's agentic workflow patterns for babysitter process orchestration
• **Implementation scope**: packages/sdk/src/runtime/, workflow definition language, GitHub integration

Implicit Procedural Knowledge

- Source: error-recovery-patterns and error-pattern-safety skills - Placement: specializations/shared/ - Why codify: Provides reusable framework for robust error handling across all types of automation - Sketch: Error detection -> Classification (transient/non-transient) -> Recovery strategy selection -> Circuit breaker implementation -> Monitoring and adjustment

• **Error Classification and Recovery Strategy**: Systematic approach to categorizing errors as transient vs non-transient and applying appropriate recovery patterns

- Source: Overall framework architecture and custom-agents skill - Placement: methodologies/agentic-workflows/ - Why codify: Systematic approach to agent workflow design that's applicable beyond GitHub automation - Sketch: Requirement analysis -> Natural language workflow definition -> Safety constraint identification -> Execution environment setup -> Guardrail implementation -> Testing and validation

• **Agentic Workflow Design Methodology**: Process for translating natural language requirements into executable agent workflows with safety guardrails

Library Mapping

Plugin Marketplace Mapping