agents.md

Resolving PR review comments and merging stacked PRs — gathering all feedback, addressing every finding, delegating to expert subagents, managing review states, and merging bottom-to-top

Stacked PR workflow with vanilla Git — creating and managing PR stacks, parallel stacks with git worktrees, propagating fixes up a stack, and safe merging bottom-to-top

Resolving PR review comments and merging stacked PRs — gathering all feedback, addressing every finding, delegating to expert subagents, managing review states, and merging bottom-to-top

Stacked PR workflow with vanilla Git — creating and managing PR stacks, parallel stacks with git worktrees, propagating fixes up a stack, and safe merging bottom-to-top

GitHub development workflow — issue discipline, commit format, pre-PR conformance gate, gh CLI recipes, and PR review handling

Writing well-structured GitHub issues — standard templates, removal issue discipline, design doc references, test update requirements, epic sizing, and dependency declaration

RAG data pipeline patterns — chunking strategies, vector store selection, embedding optimization, data validation, schema evolution, and evaluation metrics

ADR templates, API documentation standards, technical writing voice, changelog patterns, and code comment guidelines

Docker, Kubernetes, secrets management, IaC, CI/CD pipelines, and deployment strategies for Python + LLM applications

structlog setup, LLM monitoring, OpenTelemetry, Prometheus metrics, PII redaction, cost tracking, and dual-model observability

pytest patterns, fixtures, mocking, property-based testing, async tests, coverage analysis, and LLM-specific testing strategies

Multi-agent architectural deliberation protocol for heterogeneous model debate on complex design problems

Python CLI design with typer and rich — command structure, output formatting, progress indicators, error UX, verbosity levels, and machine-readable output patterns

Design patterns for LLM systems targeting both small local models (Ollama 4B-8B) and large cloud models (GPT-5, Claude Sonnet/Opus) — schema design, prompt adaptation, provider abstraction, testing strategy, and cost optimization

LangChain, LangGraph, LiteLLM patterns — chain composition (LCEL), structured output across model sizes, RAG, model routing, Ollama local models, and provider-agnostic design

mem0 integration patterns — when to use memory, hook-based auto-capture, scoping strategy, and practical examples per agent type

Prompt engineering techniques for dual-model systems — template structure, few-shot design, structured output prompting, model-adaptive strategies for both small (4B-8B) and large (GPT-5, Claude) models

Python 3.11+ best practices — typing, project structure, error handling, async patterns, testing with pytest, Pydantic configuration, and packaging with uv

Release pipeline patterns — python-semantic-release with conventional commits, PyPI trusted publishing, Docker multi-arch builds with GHCR, GitHub Actions workflows, and version tagging strategy