메뉴
Boot a Konjo session for lopi. Produces a Session Brief, runs Discovery, identifies the next sprint. Use at the start of any work session or when invoked with /konjo.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
Konjo sprint completion checklist and session handoff template. Use when closing out a sprint or ending a work session, in any consuming repo.
Documents the lopi CLI — a Rust multi-agent orchestrator that runs Claude Code agents concurrently in git-isolated branches. Use when the user asks to run a coding task through lopi, check on running/queued lopi agents, review lopi task history or logs, or interpret lopi's task status/output.
Retrofit the Konjo Quality Framework onto an existing repo that predates it. Use when asked to add konjo quality gates, improve code quality, audit an existing codebase, or run a quality sprint on any repo. Provides the step-by-step migration plan and triage protocol.
Full lopi project context — complete phase plan, current health metrics, standing rules for agents, crate details. Auto-load when working on lopi sprint planning, architecture decisions, or Phase 5+ work.
SOC 직업 분류 기준
| name | konjo-boot |
| description | Boot a Konjo session for lopi. Produces a Session Brief, runs Discovery, identifies the next sprint. Use at the start of any work session or when invoked with /konjo. |
| user-invocable | true |
Read in order: CLAUDE.md, README.md, CHANGELOG.md, PLAN.md, MASTER_PLAN.md, docs/ (if it exists). Do not skip. Do not assume contents.
REPO lopi — high-performance Rust agent orchestrator for Claude Code
LAST SHIPPED [most recent meaningful change from CHANGELOG.md]
OPEN WORK [stated next steps from PLAN.md]
BLOCKERS [failing tests, broken modules, open issues]
HEALTH [Green / Yellow / Red — one line]
Unknown is stated as unknown. Fabricated state is a lie to the next session.
Before executing any sprint, ask:
Search: arXiv (agent scheduling, LLM orchestration, self-improving systems), GitHub (new Claude Code tooling, agent frameworks), HuggingFace (relevant models/benchmarks).
If PLAN.md exists: load it, validate against codebase, flag drift. If drift found:
PLAN DRIFT
✗ [item] appears completed — not marked done
✗ [item] references [module] that no longer exists
CORRECTED NEXT STEP: [what actually needs to happen]
If no plan: run Discovery Protocol → propose a sprint:
PROPOSED SPRINT [N — Name]
MOTIVATION [real problem, real user, real value]
RESEARCH [papers, repos, techniques informing this sprint]
DELIVERABLES [concrete, shippable, verifiable things]
SUCCESS CRITERIA [tests, benchmarks, features working end-to-end]
SCOPE / RISKS [Small / Medium / Large — what could block this]
Small/Medium: propose and proceed. Large or irreversible: propose and confirm.
This skill activates on any of:
konjokonjo lopilopi konjoread KONJO_PROMPT.md and begin