메뉴
Use when the user explicitly invokes $dev-flow, or an already-loaded controller needs development lifecycle routing across multiple gates, resume/handoff state, or unclear ownership. Do not use for ordinary implementation or a single known leaf task.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
Use when non-trivial software engineering or an engineering completion claim involves implementation, debugging, refactoring, migration, architecture, technical planning, code review, material assumptions, scope pressure, or unverified evidence. Skip for trivial mechanical edits and pure factual explanation.
Use only when the user explicitly invokes $agent-debate, or when a loaded workflow routes here, and same-topic shared material needs adversarial product tradeoffs, requirement convergence, scope pruning, or decision clarity. Do not use for lanes, returned-lane review, code review, or single-agent brainstorming.
Use only when the user explicitly invokes $agent-eval, or when an active workflow routes here, to convert locked Spec behavior into acceptance evidence, automated/manual checks, failure conditions, fixtures, risk coverage, and implementation evidence requirements before Plan.
Use when the user explicitly invokes $agent-improve or asks what is worth improving in a codebase or branch and wants a read-only ranked audit. Return source-backed findings and stop. Do not use to review a specific artifact or diff (agent-review), implement a known change (steady-coding), or deliver a chosen finding (dev-flow).
Use only when the user explicitly invokes $agent-plan, or when an active workflow routes here, to convert locked Spec and Eval into implementation tasks, coverage matrix, dependencies, files/interfaces, verification commands, evidence expectations, risks, stop conditions, and lane candidates before execution.
Use only when the user explicitly invokes $agent-review, or when a loaded workflow routes here, for one focused reviewer or multiple agents to review the same artifact: Spec, Eval, plan, design, PR, diff, implementation, evidence package, or final result. Do not use for debate, lanes, or returned-lane integration.
| name | dev-flow |
| description | Use when the user explicitly invokes $dev-flow, or an already-loaded controller needs development lifecycle routing across multiple gates, resume/handoff state, or unclear ownership. Do not use for ordinary implementation or a single known leaf task. |
Route lifecycle to one loaded owner and continue. Own phase selection and closeout, not leaf work.
SELECT -> LOAD -> CONTINUE.
Use for explicit $dev-flow or uncertain ownership across gates, recovery,
lanes, integration, or closeout. Route an explicit single-leaf request straight
to its leaf.
| Need or state | Owner |
|---|---|
| Restore uncertain persisted state | project-context |
| Discover, converge, and baseline requirements | agent-requirements-analysis |
| Lock behavior | agent-spec |
| Define acceptance evidence | agent-eval |
| Build tasks or lane candidates | agent-plan |
| Implement, verify, repair, recheck | steady-coding |
| Review one artifact or result | agent-review |
| Resolve same-topic disagreement | agent-debate |
| Dispatch independent lanes | agent-lanes |
| Classify returned lanes / next batch | integration-review |
| Govern a decision discussion | discussion-workflows |
| Maintain durable technical truth | doc-driven-workflows |
| Enter explicit long-task automation | agent-self-driving |
| Check capability, authorization, sessions | agent-runtime |
| Close, pause, or hand off after evidence | dev-flow |
project-context.SKILL.md completely, then its conditionally required refs.An owner label is not a transition. Report unavailable children; never make the user dispatch internal workflows.
A selected agent-improve finding does not pass a lifecycle gate. Recheck its
evidence, choose the first missing gate and one owner, and continue. Do not
repeat the broad audit.
With active agent-self-driving, use leaf-route-only: retain the controller,
load a leaf, and return evidence. Never restart, nest, or route back into it.
Enter only after explicit $agent-self-driving invocation when none is active.
discovery -> convergence -> Requirements Baseline -> Spec -> Eval -> Plan -> execute -> review/test -> repair/recheck -> close
requirement -> behavior -> check -> task -> evidence
Start at the first source-evident missing gate. Lightweight uses compact chat
trace and direct execution; Standard separates confusable gates; Durable
persists recovery state; Lane adds ownership, collision, and integration.
Child contracts own gates: Spec and Plan retain mandatory whole-artifact review;
Lightweight makes it compact, not optional. Implementation verification and
repair belong to steady-coding. When execution follows an accepted Plan,
agent-review checks the result against that Plan before close unless the user
explicitly waives it.
Apply mode-gate.md before shaping artifacts. Read coverage-trace.md for confusable trace items, persistence, lanes, handoff, or close; read task-checkboxes.md when Plan/lane/handoff state tracks execution. Read critical-overrides.md before persisted writes, parallel dispatch/integration, completion claims, or authorization/user-decision boundaries. Read artifact-layout.md only when artifacts must persist for risk, handoff, multi-agent continuity, or recovery.
Mark complete only from source-matched status, current checks, rechecked
repairs, no unresolved accepted blocker/major, and every required trace item
completed or validly deferred after its owning artifact was updated.
Use paused when a true user decision, unavailable required dependency,
authorization boundary, missing access, or unavailable verification prevents a
safe next action. Record the raw blocker evidence, owner, effect, and exact
resume condition; never present paused work as complete.
Persist a handoff only when recovery or another thread/owner needs it. Otherwise load the next owner and continue.