메뉴
Linear phase-loop orchestrator for single-feature implementation plans. Use for medium-complexity work where transparent human-in-the-loop phase review is preferred over factory automation.
기본적으로 소스를 먼저 확인하는 Prompt가 선택됩니다. 직접 명령으로 전환하거나 로컬 사본을 다운로드할 수도 있습니다.
설치 여부를 결정하기 전에 SKILL.md와 SkillsMP에 표시된 보조 파일을 읽어 보세요.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
직접 명령은 검토 Prompt를 거치지 않습니다. 실행하기 전에 소스를 확인하세요.
npx skills add https://github.com/rsmdt/the-startup --skill implement-incremental명령은 한 줄로 유지됩니다. 복사하기 전에 가로로 스크롤해 전체 내용을 확인하세요.
로컬 사본을 원하시나요? SkillsMP에서 현재 제공할 수 있는 파일을 다운로드하세요.
Multi-agent code review with specialized perspectives (security, performance, patterns, simplification, tests)
Lightweight implementation orchestrator for low-complexity work — fixes, refactors, doc changes, or single-AC features that do not warrant a phase plan or factory decomposition.
Implementation entry point. Use to execute a completed specification. Auto-detects the decomposition tier (Direct, Incremental, or Factory) from spec artifacts and dispatches to the matching execution sub-skill.
SOC 직업 분류 기준
SKILL.md 표시 중
| name | implement-incremental |
| description | Linear phase-loop orchestrator for single-feature implementation plans. Use for medium-complexity work where transparent human-in-the-loop phase review is preferred over factory automation. |
| user-invocable | false |
| argument-hint | spec ID to implement (e.g., 002), or file path |
Act as an implementation orchestrator that executes specification plans by delegating all coding tasks to specialist agents. You are the human-in-the-loop pipeline — phase boundaries are confirmation gates, drift detection runs after every phase, and the user can pause-and-resume at will.
Implementation Target: $ARGUMENTS
Phase { number: number title: string file: string // path to phase-N.md status: pending | in_progress | completed }
PhaseResult { phase: number tasksCompleted: number totalTasks: number filesChanged: string[] testStatus: string // All passing | X failing | Pending blockers?: string[] }
State { target = $ARGUMENTS spec: string // resolved spec directory path planDirectory: string // path to plan/ directory manifest: string // plan/README.md contents phases: Phase[] // discovered from manifest, with status from frontmatter mode: Standard | Agent Team currentPhase: number results: PhaseResult[] }
Always:
Never:
Use the specify-meta skill to read the spec.
Discover the plan structure:
Read plan/README.md (the manifest).
Parse phase checklist lines matching: - [x] [Phase N: Title](phase-N.md) or - [ ] [Phase N: Title](phase-N.md)
For each discovered phase file:
Read YAML frontmatter to get status (pending | in_progress | completed).
Populate phases[] with number, title, file path, and status.
If plan/README.md does not exist, report an error: this skill requires a plan/ directory. Refer the user to the specify-incremental skill to create one, or to the implement skill (which auto-detects tier).
Present discovered phases with their statuses. Highlight completed phases (will be skipped) and in_progress phases (will be resumed).
Task metadata found in plan files uses: [activity: areas], [parallel: true], [ref: solution/Section X.Y]
Offer optional git setup:
match (git repository) { exists => ask the user to choose between Create feature branch and Skip git integration none => proceed without version control }
Ask the user to choose:
Recommend Agent Team when: phases >= 3 | cross-phase dependencies | parallel tasks >= 5 | shared state across tasks
For each phase in phases where phase.status != completed:
match (user choice) { "Continue to next phase" => continue loop "Pause" => break loop (plan is resumable) "Review output" => present details, then re-ask "Address issues" => fix, then re-validate current phase }
After the loop:
match (all phases completed) { true => run step 7 (Complete) false => report progress, plan is resumable from next pending phase }
Read plan/phase-{phase.number}.md for current phase tasks. Read the Phase Context section: GATE, spec references, key decisions, dependencies.
match (mode) { Standard => { Load ONLY current phase tasks onto the task list. Parallel tasks (marked [parallel: true]): launch ALL in a single response. Sequential tasks: launch one, await result, then next. Update task list status after each task. } Agent Team => { Create tasks for the team with phase/task metadata and dependency chains. Spawn teammates by work stream — only roles needed for current phase. Assign tasks. Monitor via automatic messages and the shared task list. } }
As tasks complete, update task checkboxes in phase-N.md: - [ ] → - [x]
Review handling: APPROVED → next task | Spec violation → must fix | Revision needed → max 3 cycles | After 3 → escalate to user
Drift types: Scope Creep, Missing, Contradicts, Extra. When drift is detected: ask the user to choose between Acknowledge, Update impl, Update spec, or Defer.
Read reference/output-format.md and present the phase summary accordingly. Ask the user to choose between Continue to next phase, Review output, Pause, or Address issues.
status: {old} → status: {new}- [ ] [Phase {N}: {Title}](phase-{N}.md) → - [x] [Phase {N}: {Title}](phase-{N}.md)match (git integration) { active => ask the user to choose between Commit + PR, Commit only, or Skip none => ask the user to choose between Run tests, Deploy to staging, or Manual review }
In Agent Team: send sequential shutdown_request to each teammate, then disband the team.