Menu
当用户需要在长会话的逻辑边界手动压缩上下文、做 context engineering、目标复述、上下文审计、外部 scratchpad、handoff 或 compaction policy 时使用。
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
| name | strategic-compact |
| description | 当用户需要在长会话的逻辑边界手动压缩上下文、做 context engineering、目标复述、上下文审计、外部 scratchpad、handoff 或 compaction policy 时使用。 |
在长会话中，上下文窗口有限且会腐烂。此 skill 指导何时压缩、保留什么、丢弃什么，以及如何把 durable state 外部化。
核心原则：在逻辑边界压缩，不在任意时刻压缩。
在重大阶段开始、handoff、压缩前，用一段短文本执行：
objective:
plan_5_steps_or_less:
context_audit:
- keep:
- externalize:
- discard:
- stale_or_conflicting:
next_boundary:
计划最多 5 步。若无法用 5 步表达，先收敛当前 tranche，不要扩大范围。
| 当前阶段 | 下一阶段 | 是否压缩 | 理由 |
|---|---|---|---|
| 研究/探索 | 规划 | 是 | 探索细节不需要带入规划 |
| 规划 | 实现 | 是 | 保留计划，丢弃规划过程 |
| 实现步骤 N | 实现步骤 N+1 | 否 | 实现中途压缩会丢失上下文 |
| 实现完成 | 验证 | 可选 | 如果上下文接近上限 |
| 验证 | 提交 | 否 | 验证结果需要带入提交 |
| 任务 A 完成 | 任务 B 开始 | 是 | 不同任务间压缩 |
压缩前必须分类：
外部化优先使用 repo artifact、checkpoint、issue/PR 描述、run log 或用户指定 scratchpad。不要把 raw session logs 当成当前事实来源。
必须保留：
可以丢弃：
当感觉上下文即将耗尽时：
Guard long, ambiguous, or stateful AI-agent work from drift. Use when the user asks to run or continue a multi-step task, autonomous loop, bug fix, repo change, PR readiness check, compaction handoff, resume from previous context, cost-control checkpoint, or any task likely to span many tool calls, files, sessions, agents, or verification gates.
Use before an agent-produced diff is committed, pushed, opened as a PR, merged, landed, or applied to user files when explicit implementation approval is missing. Trigger for review gate, review pack, approve before landing, diff first then land, human approval, merge gate, commit gate, push gate, or PR readiness. Produces a concise review pack, records risks and verification, and blocks landing until a human explicitly approves or approves with required fixes.
Audit, design, categorize, distribute, and measure agent skills using lessons from Anthropic's Lessons from building Claude Code: How we use skills. Use when reviewing an existing skill, deciding whether a workflow deserves a skill, planning a skill library, turning team knowledge into skills, choosing skill categories, writing trigger descriptions, designing progressive disclosure, or planning skill marketplace and usage measurement.
Create new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, update or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy.
Use when a skill is brittle, drifting, repeatedly failing, or needs a Reliable Skill Contract. Trigger for phrases like skill lifeguard, reliable skill, self-maintaining skill, negative examples, verification checkpoints, drift signals, replay hooks, or failure log to skill patch. Audits or patches skills so high-value workflows include explicit forbidden behaviors, checkpoints, machine-checkable done conditions, replay or smoke hooks, and drift detection.
回顾最近 N 天的 Claude Code 使用记录——扫描原始会话数据，按主题分组汇总"我都做了什么"，并从个人操作系统视角输出模式、风险与增删建议。当用户说 /recap、"看看我这几天做了什么"、"回顾一下我最近的会话"、"这两天我用 claude 干了啥"、"活动回顾" 时使用。