메뉴
Use at the start of work in this repository, when asked about the agent workflow itself, or when unsure which workflow-* skill to load. Routes to the right workflow skill without injecting long rules every turn.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
Use when a formal work item is ready to execute, when dispatching just-demand-researcher, just-demand-coder, just-demand-tester, or just-demand-advisor subagents, or when building context for focused execution.
Load this skill first for repo work so the correct just-demand routing rules are applied before other workflow skills.
Use when the user proposes a need, request, feature, design/refactor, bug report, symptom, phenomenon, vague correction, or expected-vs-actual mismatch before implementation. This is the second-priority skill after using-just-demand, including when a conversation pivots from ordinary Q&A to a concrete request, bug, or correction.
Use when the user proposes a new goal, feature, bugfix, design, refactor, research item, UI request, or any unclear work that may become a formal work item. Shapes work into task-ready intake after socratic-clarification has handled requests, bugs, pivots from ordinary Q&A, and other material uncertainty.
Use after implementation, after just-demand-tester output, when verification fails, when the user gives correction feedback, or before claiming a workflow task is done.
Use when a formal work item is ready to execute, when dispatching just-demand-researcher, just-demand-coder, just-demand-tester, or just-demand-advisor subagents, or when building context for focused execution.
SOC 직업 분류 기준
| name | using-just-demand |
| description | Use at the start of work in this repository, when asked about the agent workflow itself, or when unsure which workflow-* skill to load. Routes to the right workflow skill without injecting long rules every turn. |
Use this repository as an OpenCode-first local agent workflow runtime.
Before shaping work, identify the current situation:
No active formal task -> use workflow-intake.
Formal task ready to execute -> use workflow-execution.
Implementation/check output needs verification -> use workflow-verification.
Durable preferences, decisions, facts, open questions, or deferred options appear -> use workflow-memory.
Do not expose internal workflow mechanics to the user unless they are explicitly designing the workflow runtime.
.just-demand/ machine state.workflow-* subagent can do it from a formal task package.| Situation | Load |
|---|---|
| User proposes or clarifies work before a formal task exists | workflow-intake |
| A formal work item is ready for execution or subagent dispatch | workflow-execution |
| Reporting completion, failed verification, or correction feedback | workflow-verification |
| Recording durable decisions, preferences, facts, open questions, or deferred options | workflow-memory |
<workflow-state> breadcrumb in main-session messages. Tasks should be inspected explicitly via list-active scripts.workflow-* subagents.python3 .just-demand/scripts/task.py --root . list-active to inspect unfinished tasks before dispatch..opencode/plugins/, .opencode/agent/, .opencode/skills/, or .opencode/package.json.python3 -m unittest tests.just_demand.test_workflow_core -vnode --test tests/just_demand/test_opencode_plugins.mjspython3 -m json.tool .opencode/package.jsonpython3 .just-demand/scripts/task.py --root . list-active