菜单
Create or update system_context.json — autonomous codebase analysis that produces a thorough system description. Reads project structure, CLAUDE.md, and key source files to synthesize product overview, architecture, and technical constraints. Use standalone or invoked by /xp-plan.
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
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.
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.
| name | xp-system-context |
| description | Create or update system_context.json — autonomous codebase analysis that produces a thorough system description. Reads project structure, CLAUDE.md, and key source files to synthesize product overview, architecture, and technical constraints. Use standalone or invoked by /xp-plan. |
| effort | high |
| context | fork |
| agent | xp-agents:xp-system-analyzer |
| allowed-tools | ["Read","Glob","Grep","Bash(*/append.sh *)","Bash(*/init.sh)","Bash(*/smm/system_context_cli.py *)"] |
!CLAUDE_PLUGIN_DATA="${CLAUDE_PLUGIN_DATA}" ${CLAUDE_SKILL_DIR}/scripts/preload.sh
This skill should run as a forked subagent (xp-system-analyzer). Your agent definition contains all instructions — follow them, analyze the codebase, write system_context.json, and then report back your summary 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-system-analyzer 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.
