exo

exo

Personal developer tooling monorepo. Centralizes reusable Claude Code skills, plugins, commands, scripts, GitHub Actions, project templates, and shared Python libraries.

Install as a Claude Code plugin

# Register the marketplace (one-time)
claude plugin marketplace add /path/to/exo

# Install the plugin
claude plugin install exo@exo

This makes all skills and commands available in any project. To update after pulling changes:

claude plugin marketplace update exo
claude plugin update exo@exo

Repository structure

Directory	Purpose
skills/	Claude Code skills (SKILL.md per skill)
commands/	Claude Code slash commands (.md files)
agents/	Claude Code agent definitions (.md files)
hooks/	Claude Code hooks (hooks.json)
.github/workflows/	Reusable GitHub Actions workflows
.github/actions/	Composite GitHub Actions
templates/	Project templates (copier format)
scripts/	Standalone utility scripts
lib/	Shared Python library (exo)

Skills

Skills are self-contained capabilities that can be invoked independently or composed by agents. Each skill lives in skills/<name>/SKILL.md.

Dev-loop pipeline

The dev-loop agent orchestrates these skills into an end-to-end issue-to-PR pipeline. Each skill can also be used standalone.

Skill	Step	Description
issue-pick	1	Find, claim, and complete GitHub issues labeled for automated development
nano-spec	2	Create and manage lightweight task specifications (README, todo, doc, log)
branch-pr	3	Create a git branch and draft PR from issue context
bdd-author	4, 8	Write BDD feature files from issue content and generate step definitions
ui-design	5	Detect if a feature needs UI work and orchestrate superdesign generation
review-gate	Gate	Post plan and BDD spec as a PR comment for human review
implement-phases	6	Execute nano-spec phases with research, coding, JIT testing, and commits
jit-test	6c	Generate bespoke tests for pending changes; promote durable tests to permanent suite
ui-verify	7	Static template audit + visual fidelity comparison against designs
test-loop	9, 10	BDD test fix loop and full test suite with regression fixes
code-review	11	Self-review own PR or review any PR by number
showboat-proof	12	Generate executable proof documents demonstrating features work
submit-pr	13	Finalize PR with traceability, test results, and proof links

Other skills

Skill	Description
mutate	Mutation testing — validate test quality by checking if tests catch injected faults
req-decompose	Decompose requirements into implementable GitHub issues

Agents

Agents are autonomous multi-step workflows. Skills do the work; agents coordinate them. Each agent lives in agents/<name>.md.

Agent	Description
dev-loop	End-to-end issue-to-PR pipeline — picks an issue, plans, implements, tests, reviews, and submits
self-review	Reviews a PR diff for security, readability, architecture, and best practices
jit-test	Generates bespoke tests for pending code changes before commit

The self-review and jit-test agents are also available as skills (code-review and jit-test) for use in the dev-loop pipeline and standalone invocation.

Commands

Commands are slash-command entry points that parse arguments and delegate to agents or skills.

Command	Description
/dev-loop	Autonomous issue-to-PR loop with optional --review-plan for human checkpoints
/req-decompose	Decompose requirements into GitHub issues (scan, create, status)
/ui-audit	Run static template analysis for design system compliance
/showboat-proof	Generate executable proof documents

Dev-loop workflow

The dev-loop agent drives features from issue to PR in 13 steps, with every step backed by a standalone skill: