一键导入
attempt-artifact-consistency
Check that worker reports, commit docs, job status, git history, changed-file lists, and test logs describe the same worker attempt.
用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
菜单
Check that worker reports, commit docs, job status, git history, changed-file lists, and test logs describe the same worker attempt.
用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
| name | attempt-artifact-consistency |
| description | Check that worker reports, commit docs, job status, git history, changed-file lists, and test logs describe the same worker attempt. |
Use when reviewing or retrying a worker attempt, especially after a job was blocked
for stale reports, wrong attempt numbers, wrong commit hashes, missing test
records, or contradictory uncommitted documentation.
This skill is usually triggered by the Codex supervisor from audit artifacts such
as attempt_consistency.attempt-N.md, status.json, feedback.md, and
.ai/commit_docs/. The worker may also use it before writing a final report.
status.json attempt, final commit, test exit, and report paths
against the current attempt artifacts.base_sha..HEAD or base_sha..commit for attempt diff and
commit-range checks.changed_files.attempt-N.txt, diffstat, and commit docs describe the
same changed files.*_uncommitted*, wrong-attempt, or
wrong-final-commit documentation before review.scripts/check_attempt_consistency.py when available and block reviewer
handoff if it reports issues.When reporting a consistency decision, include:
*_uncommitted.md file remains after the loop creates a fallback
commit.status.json points to one commit while commit docs or reports name another.Create a small, scoped worker job for Cursor using the filesystem job protocol.
Review scientific or numerical code written by a worker for correctness, scope, tests, and documentation.
Design focused tests for numerical algorithms, convergence behavior, analytic cases, and backend parity.
Guide implementation and review of algorithms from papers, equations, or technical references.
Review Kokkos, MPI, GPU, or performance-portable work for correctness, parity, and maintainability.