菜单
End-to-end closed-loop build flow (brief -> plan -> forge skills -> execute -> validate). Use when you want a repeatable build loop that improves itself by forging reusable skills over time.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Validate both the standard Forgeloop HUD and the spectator-facing Director Mode against the same canonical loopback state. Use when changing elixir/priv/static/ui/*, when adding stream/broadcast presentation features, or when you need to prove the HUD spectacle still reflects repo-local truth.
Create reviewable, heavy-atom git commits from current changes (working tree by default; optionally staged/unstaged/range), then push a branch and open a PR. Use when you need to split mixed changes into a small number of cohesive commits (not tiny commits) and ship via GitHub PR.
Build with rp-cli context builder plan → implement
Deep investigation with rp-cli commands: tools gather evidence, follow-up reasoning synthesizes selected context
Export a ChatGPT-ready Question / Plan / Review prompt using rp-cli
Refactoring assistant using rp-cli to analyze and improve code organization
| name | builder-loop |
| description | End-to-end closed-loop build flow (brief -> plan -> forge skills -> execute -> validate). Use when you want a repeatable build loop that improves itself by forging reusable skills over time. |
project-architect -> write/update the plan (tasks + acceptance + runbook).skillforge -> create the minimal project skills needed to execute repeatedly (under skills/<type>/<name>/)../forgeloop/bin/sync-skills.sh).completion-director -> execute one unit of work, validate, repeat.IMPLEMENTATION_PLAN.md / prd.json)skills/*/*/ (kit ships its own base skills under forgeloop/skills/)