beagle
Overview
Beagle is a Claude Code plugin marketplace, the shared repository of agent skills that every other Existential Birds tool draws on. It ships 131 skills covering code review, documentation, testing, strategy, and planning across Python, Go, Rust, Elixir, React, and iOS.
Write a skill once and it works everywhere the marketplace is installed: Daydream uses Beagle review skills as its per-stack review backends, and Amelia's Claude Code skills and commands ship through the same marketplace.
Design decisions
- Verification before verdicts. Every review skill runs a mandatory verification protocol before reporting findings. Gate 0 is anti-confabulation: the reviewer must echo the exact file, line, and code it read before issuing any verdict. Further gates check evidence, severity, and format, with per-issue-type rules: flagging an "unused variable" requires a repo-wide reference search first. The goal is fewer false positives.
- Reviews that learn. Every finding is logged with a human verdict (accept, reject, defer, acknowledge) and the skill rule that fired. A skill-improver aggregates per-rule rejection rates from those logs and emits evidence-bound edits back to the skills themselves.
- Portability as a gate. All skills conform to the Agent Skills spec; a repo check fails on any harness-specific pattern in any skill file.
- Coherent per-stack suites. Review, docs, and testing skills within a stack share output formats and load the same verification protocol. Reference files are distilled from primary sources: the Rust review checks were built chapter by chapter from Rust for Rustaceans.
Beyond code review
The marketplace also covers writing and thinking work:
- Documentation.
beagle-docsorganizes writing around Diátaxis doc types:draft-docsgenerates first drafts from code analysis,improve-docclassifies a doc's sections by type and refines each one interactively, andensure-docsfinds coverage gaps.review-ai-writingdetects robotic AI prose in docs, commit messages, and comments;humanize-beaglerewrites it. - Strategy.
strategy-interviewbuilds a strategy through guided conversation;strategy-reviewstress-tests an existing strategy document across seven dimensions, from diagnosis quality to falsifiability. - Idea to plan.
prfaq-beaglepressure-tests a concept with Amazon's Working Backwards format and returns a pass/fail verdict. On pass,brainstorm-beagleshapes the idea into a WHAT/WHY spec through structured dialogue, andwrite-planturns the spec into a TDD implementation plan of bite-sized steps.
Install
Add the marketplace, then install the plugins you need:
claude plugin marketplace add https://github.com/existential-birds/beagle
claude plugin install beagle-elixir@existential-birds
Verify the install by opening a new Claude Code session and running one of the
plugin's skills, for example /beagle-elixir:review-elixir. If the skill
loads, the plugin is active.
To use the skills with other AI agents such as Codex, install through the skills CLI:
npx skills add existential-birds/beagle
Per-plugin skill catalogs and setup details live in the repo README.
Plugin catalog
Skills are grouped into ten domain-namespaced plugins:
| Plugin | Skills | Category |
|---|---|---|
| beagle-core | 19 | Shared workflows, verification, git, skill tooling |
| beagle-python | 7 | Python, FastAPI, SQLAlchemy, pytest |
| beagle-go | 13 | Go, BubbleTea, Wish SSH, Prometheus |
| beagle-elixir | 11 | Elixir, Phoenix, LiveView, ExUnit, ExDoc |
| beagle-ios | 16 | Swift, SwiftUI, SwiftData, iOS frameworks |
| beagle-react | 28 | React, React Flow, shadcn/ui, Tailwind, Remix v2 |
| beagle-rust | 12 | Rust, tokio, axum, sqlx, serde |
| beagle-docs | 10 | Documentation quality, Diátaxis doc types, AI writing detection |
| beagle-analysis | 13 | Brainstorming, ADRs, strategy, LLM-as-judge, spec gaps, TDD plans |
| beagle-testing | 2 | Test plan generation and execution |
| Total | 131 | — |
Each plugin exposes canonical entry-point skills, such as review-elixir,
review-frontend, and review-go for stack-specific code review.