메뉴
Repository-specific workflow for Graft ai-plan governance. Use when a docs/automation task changes ai-plan router documents, active-topic recovery materials, templates, catalog coverage, or bounded ai-plan validators without creating a second startup, validation, or recovery path.
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
| name | graft-ai-plan-governance |
| description | Repository-specific workflow for Graft ai-plan governance. Use when a docs/automation task changes ai-plan router documents, active-topic recovery materials, templates, catalog coverage, or bounded ai-plan validators without creating a second startup, validation, or recovery path. |
Use this skill when a docs/automation task changes the governance shape of ai-plan/**.
Treat root AGENTS.md and ai-plan/AGENTS.md as the governance sources. This skill does not replace startup,
validation, commit, closeout, or runtime authority.
AGENTS.md startup preflight.ai-plan/AGENTS.md..ai/environment/tools.ai.yaml.ai-plan/README.md.ai-plan/design/governance/ai/AI任务追踪与恢复设计.md.ai-plan/design/governance/ai/AI工具与MCP接入治理规范.md.ai-plan/public/README.md after startup preflight.README.md, startup-prompt.md, tracking, and trace files..agents/skills/** or scripts/**, pair this skill with graft-ai-governance-audit.Use this skill when the task changes any of:
ai-plan/AGENTS.mdai-plan/README.mdai-plan/catalog.jsonai-plan/design/** documents that define ai-plan governance or ADR convergenceai-plan/public/README.mdai-plan/public/<topic>/** recovery materialsai-plan/templates/**Work Intake, Work Contract, or contract-driven bootstrap governance under ai-plan/**docs/automation validators under scripts/** that guard ai-plan governance structure.agents/skills/** only when they are part of the same ai-plan governance sliceDo not use this skill for:
server, web, OpenAPI, or runtime implementation workai-plan/** retrofit, frontmatter rollout, or projection-model expansionai-plan/**ai-plan governance; use
graft-ai-governance-audit for thatAGENTS.md as the only startup-governance sourceai-plan owner: design/, design/decisions/, roadmap/, public/, public/archive/,
lessons/, or templates/ai-plan/catalog.json as supplementary, single-file, and boundedWork Intake as the only approved entry workflow for new long-running workdesign, roadmap, topic, and ADR as artifacts selected by Work Contract, not peer intake pathsai-plan/** only when authority discovery proves it is requiredai-plan/public/README.mdstartup-prompt.md aligned with root startup governance; do not invent a second receipt formatWork Intake, keep the persisted Work Contract in tracking rather than inventing a new
topic-level metadata filedocs/automation guards, not runtime or CI substitutesFor ai-plan/** governance work, always run:
git diff --check
Add:
python3 scripts/validate_ai_plan_structure.py
when the slice touches ai-plan/catalog.json, ai-plan/public/README.md, ai-plan/templates/**, active-topic
recovery structure, or the ai-plan structure guard itself.
Add:
python3 scripts/validate_ai_governance.py
when the slice touches .agents/skills/**, scripts/**, or
ai-plan/design/governance/ai/AI工具与MCP接入治理规范.md.
Run:
python3 -m unittest scripts/test_validate_ai_plan_structure.py
only when scripts/validate_ai_plan_structure.py or its tests change.
graft-work-intake own independent intake rules outside root AGENTS.md, ai-plan/AGENTS.md, and
the approved ADRscompose-project-management implementation from ai-plan governance workai-plan linterReport:
ai_plan_governance:
- authority_owner:
- router_paths_changed:
- concurrent_topics_preserved:
- validators_run:
- scope_expanded:
Repository-specific loop orchestrator for Graft multi-agent tasks. Use when one bounded task should run through repeated same-session serial worker-subagent rounds of `graft-multi-agent-task` under the default `topic-completion-loop` mode until the topic reaches `archive-ready`, becomes `blocked`, or no remaining batches stay in scope.
Repository-specific multi-agent coordination workflow for Graft. Use when the user explicitly wants subagent delegation, or when the current task cleanly splits into two or more disjoint slices across server, web, docs, or automation, and the main agent should keep ownership of review, validation, and final integration. `graft-boot` should assess whether this skill is justified before delegation starts.
Repository-specific workflow wrapper for Graft multi-agent tasks. Use when a task should be executed through `graft-multi-agent-batch`, and the slice may need `graft-task-closeout` plus `graft-commit` to finish with a safe handoff after validation.
Repository-specific end-of-task closeout workflow for Graft. Use when a task slice is ending and the agent needs to decide whether to emit only a next-step startup prompt or to commit the current validated slice first and then hand off safely. This is the default slice-end path for work started through `graft-boot`.
Manage reusable Graft AI-agent worktrees through guarded status, acquire, and release commands without treating directories or task branches as long-lived ownership.
Repository-specific Playwright browser workflow for inspecting, authenticating into, and interacting with the local Graft web UI. Use when Codex needs Graft browser screenshots, DOM text snapshots, login/auth verification with temp credentials, or simple click/fill/wait checks before normal web validation.