菜单
Review the completed sprint. Compares what shipped vs planned, updates execution_plan.json milestones with delivered status, and records sprint velocity. Use when all stories are done or deferred.
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 | xp-sprint-review |
| description | Review the completed sprint. Compares what shipped vs planned, updates execution_plan.json milestones with delivered status, and records sprint velocity. Use when all stories are done or deferred. |
| effort | high |
| context | fork |
| agent | xp-agents:xp-sprint-reviewer |
| allowed-tools | ["Bash(*/append.sh *)","Bash(*/init.sh)","Bash(*/skills/*/scripts/*)","Bash(python3 */smm/plan_cli.py *)"] |
!CLAUDE_PLUGIN_DATA="${CLAUDE_PLUGIN_DATA}" ${CLAUDE_SKILL_DIR}/scripts/preload.sh
This skill should run as a forked subagent (xp-sprint-reviewer). Your agent definition contains all instructions — follow them, record the result, and then report back your full findings to the main agent.
If you are the main agent and see this: do not do this work yourself. This skill must run as the xp-sprint-reviewer subagent. The subagent result is returned as a tool result which is NOT visible to the user — you must output the key findings as text in your response.
Part of the reviewer's job is to rerun every verify-bearing acceptance item across the sprint (verify_acceptance.py --sprint) before /xp-sprint-close and return the per-surface PASS/FAIL matrix — surface that matrix to the user alongside the findings, since the tool result is invisible to them. The rerun emits the deterministic verify event the close gate reads, so it must run before the close dispatch.
After you have surfaced the reviewer's findings and the verify matrix to the user, invoke /xp-sprint-close to push the sprint branch, fork the xp-close-reviewer for cross-cutting review, and merge into the target with cleanup.
Create sprint.json from an execution plan milestone or product spec. Decomposes milestones into context-rich stories with file domains, interface contracts, and inlined design context. Deep codebase dive identifies story boundaries. Customer confirms scope before writing.
The universal per-story execution-shape decision. Reads the session default tier and the plan-reviewer's per-story recommendation, then applies a behavior table: it may exit in-agent (no spawn — continue in the existing checkout) or spawn exactly ONE teammate at the chosen tier (the story's executor_model forwarded to spawn_teammate's --model flag). Valid for solo and parallel frontiers alike. Runs once per story after that story's plan is reviewed.
Push the sprint branch, fork xp-close-reviewer for cross-cutting review, and merge into the target with cleanup. Auto-invoked at the end of /xp-sprint-review; degrades gracefully when gh is unavailable.
Verify acceptance criteria for in-progress stories. Present criteria, guide e2e test execution, mark stories done or deferred, update sprint.json.
This skill should be used when the user needs to select work for the session, review retro Try items, triage open questions, or pick sprint stories. Invoked as part of /xp-kickoff or standalone via /xp-work-selection.
Review the current plan mode plan. Checks plan size, TDD ordering, milestone boundaries, and decision conflicts. Use after planning completes.
