Browse / General-purpose / agent-workspace-template
agent-workspace-template
punokawan
One workspace for running several product projects with Claude Code and Codex from a single shared toolkit: 27 specialist agents, 35 slash commands, 59 skills and the scripts…
Open punokawan/agent-workspace-template → Checked against the four rules.
Works with any agent
ships AGENTS.md, the cross-tool instruction file.
In its own words
One workspace for running several product projects with Claude Code and Codex from a single shared toolkit: 27 specialist agents, 35 slash commands, 59 skills and the scripts behind them, plus the marketing-skills plugin (61 skills, installable in Claude Code and Codex). Each project picks how strict its rules are.
What you get
- One guide per level.
AGENTS.mdat the workspace, project and repo level; eachCLAUDE.md
only imports its AGENTS.md, so both agents read the same text. The guide nearest the edited file wins.
- Rule profiles per project.
light,standardorstrict, or single modules on and off.
Core rules (precedence, security, confirm before outward actions, never write production data) always apply.
- Worktree flow. Every change on a short-lived branch in
.wt/, with project guides and
skills linked in automatically.
- Verification and run state.
/verifyreports real exit codes;/checkpointand/resume
let another session pick work up.
- Project data stays in the project. Shared skills hold no project data; project and repo
skills live with their project. .agents/skills links let Codex find the same skills.
- A self-check.
/doctorfails on stale rosters, dead references, stale rule blocks and
broken links.
Quoting 28% of the README. Read the rest on GitHub ↗
Read from the repository's git tree on 2026-08-17.