Verified project record
wshobson/agents
wshobson/agents provides a single Markdown source-of-truth marketplace for agentic plugins that deploys natively across five different AI coding harnesses. It emits harness-native artifacts rather than lowest-common-denominator translations and ships granular, single-purpose installable units.
Project overview
Ships to five agentic harnesses from one Markdown source-of-truth, generating harness-native artifacts without lowest-common-denominator translations.
- Project type
- AI Agent · AI Coding · Prompt Engineering
- Use cases
- Coding & Development
- Deployment
- Refer to project documentation
- License
- MIT
Best for
- Developers and AI engineers who work across multiple agentic coding harnesses and prefer to manage plugin configurations from a single Markdown source.
Key capabilities
- Ships granular, single-purpose installable units including agents, skills, commands, and orchestrators.
- Emits harness-native artifacts for five different agentic harnesses from a single Markdown source rather than lowest-common-denominator translations.
- Provides a three-layer evaluation framework (static, LLM judge, Monte Carlo) for measuring and certifying plugin/skill quality. The LLM judge layer relies on external models for inference, making external inference an optional dependency for using this specific evaluation capability.
- Includes Pensyve as an external git-subdir entry for Claude Code.
Limitations and risks
- Transformed trees for Gemini and OpenCode are gitignored and require generation via make commands during setup or post-install operation.
- The data boundary is evaluated as external_optional. The evaluation framework relies on external models for its LLM judge component, meaning operations utilizing this layer require remote inference.
- Opt-in tiered models like Fable 5 imply premium costs during inference. The project recommends various Claude models for tiers of tasks, making paid services optional.
Getting started
- Setup difficulty is medium, requiring cloning and running make commands for some harnesses. Plugin installation syntax varies per harness. The first success path is to run `/plugin marketplace add wshobson/agents` followed by `/plugin install python-development`.
Alternatives and comparisons
- Provides specialized, personality-driven, deliverable-focused AI agents that can be installed into multiple agentic coding tools.
- Architecturally enforces divergent ideation across isolated cognitive frames and prunes traps to deepen ideas for coding agents.
- Replaces fear-based AI agent prompts with a trust-driven methodology for agents that verify their work.
Project comparisons
Evidence and sources
- GitHub project description: Multi-harness agentic plugin marketplace for Claude Code, Codex CLI, Cursor, OpenCode, GitHub Copilot, and Gemini CLI
- README: > Production-ready agentic workflow building blocks: **94 plugins**, **203 agents**, > **175 skills**, **109 commands** — built for Claude Code and consumed natively by > OpenAI C…
- README: /plugin marketplace add wshobson/agents /plugin install python-development # or any of 94 plugins
- README: gh repo clone wshobson/agents ~/agents && cd ~/agents make generate HARNESS=gemini && gemini extensions install . # Gemini make install-opencode # OpenCode (runs generate + symlin…
- README: **Orchestrators** | 16 | Multi-agent coordination workflows (full-stack, security, ML, incident response)
AI Search
Find projects, verify facts, compare options, or turn a complex need into an actionable plan
Try a searchA click only fills the search box; you stay in control
Project Details
0