Five GitHub Repositories Add Reusable Workflows to AI Coding Agents

AI coding agents can produce code, but engineering work also depends on requirements, tests, debugging, review and decisions about when a change is ready. Reusable agent skills package those practices as instructions and workflows, so teams can apply them without restating the same process in every session.

Five open-source collections take distinct approaches: Superpowers, Skills for Real Engineers, ECC, Awesome LLM Apps and Agent Skills. Their GitHub star counts in the source are 285.5K, 260.1K, 256.7K, 137.3K and 93.6K, respectively. The useful question is less which collection is biggest than how much process and tooling a project needs.

Workflows that shape the development process

Superpowers, created by Jesse Vincent and the Prime Radiant team, treats skills as a development methodology. Its collection covers brainstorming, implementation planning, test-driven development, debugging, Git worktrees, parallel agents, code review and verification.

For a feature request, the workflow starts by clarifying the desired change. Once the design is approved, the agent makes a plan, implements it through subagents or plan execution, follows TDD, reviews the result and verifies it before finishing the branch. Its debugging skill also calls for a root-cause process rather than repeated speculative edits. That structure could suit teams that want consistent steps across many tasks, though it means adopting a broader method rather than selecting only a single prompt.

The source gives Claude Code’s installation command as /plugin install superpowers@claude-plugins-official. It also lists Codex, Cursor, Gemini CLI, OpenCode, GitHub Copilot CLI, Kimi Code, Antigravity and other coding environments as supported.

Skills for Real Engineers takes a more modular route. Matt Pocock’s collection includes skills for interviewing a user about a change, documenting a shared domain model, turning ideas into specifications and tickets, implementation, TDD, bug diagnosis, research, code review and architecture.

One example, grill-with-docs, asks questions to clarify a proposed change while refining terminology and updating project documents such as CONTEXT.md and ADRs. The tdd skill follows red, green and refactor; diagnosing-bugs provides a structured debugging routine. For larger work, the repository describes a chain from grill-with-docs through specification, tickets and implementation to code review.

Install with npx skills@latest add mattpocock/skills. The installer lets users choose skills and coding agents, then copies files into the project for editing. Pocock also recommends running setup-matt-pocock-skills once per repository. This collection’s small, composable units may fit teams that want to assemble their own process.

Broader toolkits and working examples

ECC by Affaan Mustafa combines skills with specialized agents, rules, hooks, memory, continuous learning, security tools and integrations. The repository lists 292 skills, 68 specialized agents and 94 command shims, along with memory across sessions, continuous learning and AgentShield scanning.

Its subjects range from TDD and research to frontend development, machine learning, security, operations, architecture, testing and agent orchestration. The workflow runs from planning and testing through implementation, review and verification, then remembering and improving. Those last steps are intended to preserve useful context and turn repeated patterns into reusable skills. AgentShield scans configurations for secrets, permissions, hooks, MCP settings and prompt-related security issues.

ECC’s setup commands are npx [email protected] setup and npx [email protected] install --guided. Claude Code has the strongest integration described in the source; support for Codex, Cursor, OpenCode, Kimi Code, Gemini and other agents varies. ECC’s breadth may be excessive for someone seeking only a debugging routine, but its memory and specialized workers make it the most platform-like option in this group.

Awesome LLM Apps, by Shubham Saboo, is broader than a skills library. It contains more than 100 open-source AI agents, agent skills and retrieval-augmented generation applications using models including GPT, Claude, Gemini, DeepSeek, Llama and Qwen. Its value is in showing how models, tools, retrieval and skills can fit into working applications.

Its skills include Project Graveyard, which examines local Git repositories to identify abandoned projects and suggest one to resume; Scope Creep Detector, which checks whether a code change has expanded beyond its initial goal; and Commit Archaeologist, which uses Git history to explain why code exists. Other examples test and improve skills or divide tasks among an orchestrator, advisor and workers. Project Graveyard can be added with npx skills add https://github.com/Shubhamsaboo/awesome-llm-apps/tree/main/agent_skills/project-graveyard.

Production checks and choosing a fit

Agent Skills by Addy Osmani organizes 24 skills around defining, planning, building, verifying, reviewing and shipping software. Topics include requirements, TDD, frontend work, API design, debugging, security, performance, CI/CD, observability and documentation. Commands such as /spec, /plan, /build, /test, /review and /ship load workflows for each stage.

A notable feature is its use of quality gates. The constraint-driven-development skill lets a team set requirements for testing, linting, security, coverage and performance, then schedule more expensive checks later in the workflow. Install everything with npx skills add addyosmani/agent-skills, list skills with npx skills add addyosmani/agent-skills --list, or add one with npx skills add addyosmani/agent-skills --skill test-driven-development. The Skills CLI supports more than 70 coding agents, including Claude Code, Cursor, Codex, Copilot and Cline.

The collections span different needs: Superpowers supplies an end-to-end method, Pocock’s skills are modular, ECC bundles agents and memory, Awesome LLM Apps offers examples to explore, and Osmani’s set emphasizes production checks. A practical starting point is to choose the workflows that match a real project, try them on actual tasks and edit the SKILL.md files as the team learns what works. The payoff depends on whether those instructions make the agent’s work more consistent and easier to verify.

Source author: Abid Ali Awan (@1abidaliawan).

More on this topic

• 10 GitHub Repositories for Web Development in Python
• 10 GitHub Repositories for Modern Database Systems and Tools
• 10 GitHub Repositories to Master Quant Trading
• 10 GitHub Repositories to Master Self-Hosting
• 10 GitHub Repositories to Master FastAPI
• 5 GitHub Repositories to Learn Quantum Machine Learning