본문으로 건너뛰기
cccskills
무료GitHub에 공개

implementation-kickoff

Start or resume a requested Guardrails implementation or PR takeover in the selected linked worktree with bounded scope and verification.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

설치 방법 보기

포함된 파일(2)

  • SKILL.md4.0 KB
  • agents/openai.yaml125 B

SKILL.md(원문)

설치하기 전에 에이전트에게 전달되는 지시 내용을 확인할 수 있습니다.

Implementation Kickoff

Start from the user's actual implementation request. An investigation or a plan alone is not permission to implement, push, or publish. Preserve authorization already given; this skill adds none. Apply AGENTS.md and implementation-strategy where applicable.

Establish the checkout and scope

Record the original outcome, acceptance criteria, affected paths, non-goals, and intended base. Reuse the current task's selected linked worktree. If running in a primary checkout, follow AGENTS.md and ask to start or hand off to a worktree. Do not automatically create another worktree or change unrelated checkouts.

Refresh remote references and record the full intended starting SHA. For a new independent task use the refreshed default branch unless the user chose another base. For an explicitly requested stack, use the exact previous slice's tip. Verify the checkout equals that SHA before editing a new task. For a resumed task, preserve its existing changes and record HEAD plus the original comparison base instead of resetting it to main. Stop on an unexplained base mismatch.

Inspect tracked and untracked state before edits. Keep task deliverables distinct from existing user work and operational plans/review notes. Use codex/ branch names by default, preserving an explicitly selected branch or user preference.

Take over an existing PR

Read current metadata, author commits, discussions, and the complete PR diff. Confirm the PR is still an appropriate source. Preserve its underlying outcome, but evaluate the proposal against the current architecture and toolchain. Import only the accepted scope; a takeover does not require copying the entire patch or preserving intermediate commits.

Record the source PR and verified author/coauthor identities. Credit adapted work with valid Co-authored-by trailers, without inventing identities. If the identity needed for attribution cannot be verified, resolve it before committing. For a split takeover, describe each slice as part of the replacement effort; do not claim the first slice completes the whole source PR.

Implement and verify

Keep changes within the agreed scope. Preserve user work and avoid automatic stashing, rebasing, or rewriting unrelated branches. Check integration with an advanced target before submission; if an update is needed, handle only this task's branch and rerun checks/review affected by any content change. Do not force a detached checkout, one-commit topology, or history rewrite to manufacture a clean handoff.

Use verification and final review before pushing or creating or updating a PR. Follow the release guide: ordinary user-visible SDK changes need a changeset; internal workflow changes need none. Leave version and changelog generation to the release PR.

Handoff

Use pr-draft-summary for the final description and authorized submission/follow-up.

Inspect status, full diff statistics, and every task commit before staging or reporting. Stage only task-owned deliverables; keep local planning and review artifacts out of the PR. Preserve hooks and inspect the committed result.

Report branch, base, commit, requested outcome, verification, review status, and remaining blockers. When authorized to submit, follow AGENTS.md for current-head CI monitoring, scoped fixes, root-channel review requests, and feedback replies. For a requested stack, create each PR against its predecessor and verify that relationship. Do not assume all workflows run on feature-branch PR targets: inspect triggers and use an authorized manual CI run on each head when needed. Never merge or close the source PR without authorization.

리뷰

아직 리뷰가 없습니다. 사용해 본 소감을 남겨 주세요.

댓글

로그인 없이 쓸 수 있습니다. 사용 후기, 질문, 작동하지 않은 환경 등을 편하게 남겨 주세요.

아직 댓글이 없습니다. 첫 댓글을 남겨 보세요.

보내면 바로 공개됩니다. 이메일 주소나 전화번호 같은 개인 정보는 쓰지 마세요. URL은 2개까지입니다.

0 / 1,000자

이 양식은 reCAPTCHA로 보호되며 Google 개인정보처리방침 및 서비스 약관이 적용됩니다.

같은 저장소의 스킬

Select and run Guardrails verification for code, dependency, test, tooling, documentation, or repository-workflow changes.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

openai/openai-guardrails-js1082026년 10월 19일 업데이트

Prepare independent review of a complete Guardrails change before pushing or updating a PR, using the repository adversarial-review procedure.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

openai/openai-guardrails-js1082026년 10월 19일 업데이트

Choose bounded implementation scope and existing owning modules before Guardrails runtime, configuration, client, resource, streaming, Agents, or SDK compatibility changes and feedback fixes.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

openai/openai-guardrails-js1082026년 10월 19일 업데이트

Draft a Guardrails PR from its complete final diff and carry out authorized CI and review follow-up, including stacked PRs and takeovers.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

openai/openai-guardrails-js1082026년 10월 19일 업데이트

openai의 스킬 모두 보기

이 스킬의 문제 신고