菜单
Analyse completed pipeline runs and surface DOD pass rates, loop-backs, and failure patterns.
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
Show the Cadenza command map — a quick reference to all available agents and commands.
Turn raw context, a thread, or a rough note into a well-structured GitHub issue.
Generate a PO-ready grouped changelog from merged PRs since the last release.
Generate a PR description for the current branch using commits, diffs, and the project template.
Write PHPUnit tests for PHP source files, auto-detecting conventions from the project.
Analyse changes, create atomic commits with generated messages, then offer to push. Accepts inline hints — "without <file>", "excluding <path>", "only <path>".
| name | retrospective |
| description | Analyse completed pipeline runs and surface DOD pass rates, loop-backs, and failure patterns. |
Closes the pipeline feedback loop. Scans completed issue runs, surfaces patterns the
team should act on, and proposes ready-to-paste Section 13 learnings for AGENTS.md.
Project identity is auto-detected (see AGENTS.md §3) — no config file.
This skill uses: {TEMP_ROOT}, {REPO}, {DISPLAY_NAME}.
Parse arguments (if any):
{TEMP_ROOT}/issues/2026-05-01 2026-05-31) → filter runs by tasks.json creation dateSpawn retrospective-agent with {TEMP_ROOT}, {REPO}, {DISPLAY_NAME}, and the
date range (or "all" if not specified).
Return the output file path and a one-paragraph summary of the top finding.