GitHub repositories

Superpowers, Skills, Oh My Zsh & Agency Agents Compared

Compare the stated purposes and captured GitHub metadata of four Shell-language projects, with source links, clear limits and practical questions for evaluating them.

AltAPIs Editorial · AI-assisted · owner-approved

Programming illustration; not a screenshot of the products discussed. File:Programming code.jpg by Martin Vorel · CC BY-SA 4.0. Wikimedia thumbnail resize only; image remains under its original license.

Reported purposes and basic metadata

All four records were observed on 2026-09-19T06:35:12.688Z, with firstSeenAt dates ranging from 2026-09-15T06:33:57.641Z (obra/superpowers) to 2026-09-15T06:34:01.592Z for the other three. The obra/superpowers repository describes itself as “An agentic skills framework & software development methodology that works.” It shows 288,598 stars, 25,815 forks, lists Shell as its language, and an MIT license. The mattpocock/skills project is labeled “Skills for Real Engineers. Straight from my .agents directory.” It has 265,390 stars, 22,378 forks, Shell language, and MIT license. The ohmyzsh/ohmyzsh repository presents itself as “🙃 A delightful community-driven (with 2,500+ contributors) framework for managing your zsh configuration. Includes 300+ optional plugins (rails, git, macOS, hub, docker, homebrew, node, php, python, etc), 140+ themes to spice up your morning, and an auto-update tool.” It records 189,795 stars, 27,077 forks, Shell language, and MIT license. The msitarzewski/agency-agents description reads “A complete AI agency at your fingertips - From frontend wizards to Reddit community ninjas, from whimsy injectors to reality checkers. Each agent is a specialized expert with personality, processes, and proven deliverables.” It shows 153,428 stars, 24,751 forks, Shell language, and MIT license. These figures are taken directly from the respective repository metadata.

Concrete differences in stated scope

Superpowers frames a methodology that couples agentic skills with software development processes, suggesting a focus on workflow integration and skill orchestration. Skills appears to be a curated library of skill definitions drawn from a personal .agents directory, emphasizing reusable, engineer‑centric capabilities without prescribing a broader process. Ohmyzsh limits its scope to enhancing the Zsh shell via plugins, themes, and an auto‑update utility, targeting interactive command‑line customization. Agency‑agents positions itself as a catalog of AI‑driven personas (“agents”) each bundled with personality traits, defined processes, and claimed deliverables, aiming to provide ready‑made expert‑like assistants for varied tasks. The descriptions thus indicate distinct audiences: development methodology (superpowers), skill repository (skills), shell enhancement (ohmyzsh), and AI agent collection (agency‑agents).

Limitations evident from the supplied metadata

The description fields do not disclose implementation details, versioning, or dependency requirements. Superpowers’s claim of being an “agentic skills framework & software development methodology” lacks any mention of concrete tooling, guides, or examples within the metadata. Skills’s reference to a personal .agents directory raises questions about the generality and portability of its contents beyond the original author’s environment. Ohmyzsh’s note of “2,500+ contributors” highlights community size but does not specify compatibility with newer Zsh releases or potential plugin conflicts. Agency‑agents’ portrayal of agents having “personality, processes, and proven deliverables” provides no insight into the underlying AI models, data sources, or licensing of those agents. Consequently, any assessment of stability, extensibility, maintenance frequency, or suitability for production use would require further inspection beyond the supplied stars, forks, language, or license fields.

Concrete next steps for deeper evaluation

A reviewer could first examine each repository’s README or documentation to verify the presence of installation instructions, usage examples, and contribution guidelines. Checking the commit history would clarify recent activity and maintenance cadence. Inspecting the directory tree would reveal whether superpowers and skills provide concrete skill definitions, scripts, or configuration files, and whether ohmyzsh offers plugin and theme directories. For agency‑agents, reviewing sample agent definitions would help determine the granularity and reproducibility of the offered AI experts. The captured license fields all say MIT; read the current license files and any separately licensed materials before deciding whether reuse fits your project. Looking for tags or releases could indicate versioned snapshots. These steps rely solely on observable repository contents and avoid assuming quality, adoption, security, or performance from star or fork counts.