codevoyant is a collection of skills (slash commands) that give AI coding agents structured workflows for planning, development, and tooling. Works with Claude Code, OpenCode, and VS Code Copilot.
Workflows
 |
spec — plan and execute complex work
Research requirements, generate proposals, create phase-by-phase implementation plans, and execute them step-by-step or hand off to a background agent. |
 |
explore — pure technical exploration
Research problem spaces, compare technical approaches and repositories, and generate decision-oriented proposals. Never writes application code. |
 |
docs — engineering documentation
Generate, update, review, and retroactively create docs from standard templates. |
 |
flow — end-to-end pipeline orchestration
Chain skill workflows into end-to-end pipelines that run sequentially. |
 |
pr — AI-powered code review
Generate professional inline review comments from a diff, address change requests, and publish a draft review. |
 |
qa — bug investigation and smoke testing
Structured bug investigation, browser-agent smoke tests, one-command issue filing to GitHub, GitLab, or Linear. |
 |
skill — build, maintain, and report skills
Scaffold new skills, iterate on existing ones, audit quality, and report issues to skill authors. |
Domains (experimental context skills for specialized engineering)
 |
plan Experimental — planning at every level
Plan a single task, a project, an initiative, or a whole product — with milestone-grouped task plans, roadmap review, and Linear sync. |
 |
pm Experimental — product roadmaps and PRDs
Phased roadmaps, per-feature PRDs, prioritization review, and Linear initiative sync. |
 |
ux Experimental — prototyping and style research
Scaffold SvelteKit prototypes, create single-file wireframe explorations, and extract styles from live sites. |
Tools
 |
changelog — retcon PR/MR commit messages and preview the next changelog |
 |
cz — show current and predicted next version using commitizen |
 |
gh — Watch Actions pipelines, fetch and post inline PR review comments, manage draft reviews |
 |
git — conventional commits with auto-formatting, safe interactive rebase, and branch/worktree setup |
 |
glab — watch CI pipelines, fetch and post inline MR discussion notes, manage draft reviews |
 |
helix — Helix editor key bindings for file navigation and selection-based workflows |
 |
linear — create Linear issues and bug reports via MCP Linear tools |
 |
mise — mise.toml authoring, task naming conventions, language-specific setup recipes |
 |
release — show current and predicted next version via semantic-release or release-it |
 |
task — detect and run tasks across mise, just, task.dev, and npm scripts |
 |
vim — Vim and Neovim key bindings for file navigation, search, and splits |
npx skills add cloudvoyant/codevoyant
npx skills add cloudvoyant/codevoyant
OpenCode's skill tool takes only a name, so args typed after /skill are dropped. Dispatcher skills degrade gracefully: with no parseable verb they ask the user what they want (numbered list on non-Claude-Code platforms) instead of silently running help. See docs/installation.md.
curl -fsSL https://raw.githubusercontent.com/cloudvoyant/codevoyant/main/scripts/install-vscode.sh | bash
# Plan and execute a feature
/spec new my-feature # explore requirements and create a plan (add --branch / --worktree to isolate work)
/spec go my-feature # execute step-by-step with review stops
/spec bg my-feature # hand off to a background agent
# Ship code
/git worktree --branch --worktree # set up a branch and/or worktree (independent)
/git commit # format → conventional commit → push
/gh ci --autofix # watch GitHub Actions, auto-fix failures and re-push
/glab ci --autofix # watch GitLab CI, auto-fix failures and re-push
# Review a PR/MR
/pr new # generate inline review comments from the diff
/pr address # pull reviewer feedback and propose fixes
# Plan engineering work
/plan plan "migrate auth to OAuth2" # milestone-grouped task plan
/plan review my-plan # capacity and risk review
# Plan product work
/pm plan quarter # draft quarterly roadmap
/pm prd "user authentication" # standalone PRD
# Build a skill
/skill new my-command # scaffold from template
/skill critique my-command # audit quality before shipping
/skill feedback spec # report an issue to skill authors
See the full documentation →
MIT © Cloudvoyant