菜单
This skill should be used when the user says "context check", "check for insights", "anything worth saving", "review session", or wants to analyze the current conversation for valuable learnings worth persisting — debugging discoveries, architectural decisions, corrections, workflow 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
This skill should be used when the user says "context save", "remember this", "save this", "note this for later", "keep this", or wants to persist any knowledge, pattern, decision, or learning from the current session to persistent memory.
Scan recent Claude session logs for missed insights (fallback when Stop hook fails). This skill should be used when the user says "context scan", "scan sessions", "check old sessions", "find missed insights", or wants to recover insights from sessions where the Stop hook did not fire. Also used by hourly cron to automatically sweep for uncaptured knowledge.
This skill should be used when the user says "work done", "context done", "task done", "finish task", "complete task", "close task", or indicates they have finished working on an active task. Summarizes task insights and distributes them to each repo's insight folder.
Search and retrieve insights from persistent memory. This skill should be used when the user says "context find", "recall", "what do I know about", "find my notes on", "search memory", "do I have anything on", or wants to retrieve previously saved knowledge. Always invoke to check memory before answering any topic-based question.
This skill should be used when the user says "research", "look into", "look up", "find out about", "search for", or asks about a topic needing investigation. Checks persistent memory first, performs web research if insufficient, persists results.
| name | context-check |
| description | This skill should be used when the user says "context check", "check for insights", "anything worth saving", "review session", or wants to analyze the current conversation for valuable learnings worth persisting — debugging discoveries, architectural decisions, corrections, workflow patterns. |
Analyze the recent conversation to detect facts worth persisting.
/context check
Scan the last 10 human/assistant messages for concrete facts in these categories:
insight — completed work:
agent_edit — AI behavior changes:
task — ONLY unstarted intentions:
Good facts (concise, high-level, with WHY):
Bad facts (skip these):
context-save skill — do NOT ask for confirmationRead insights_root from ~/.claude/memory-keeper.local.md YAML frontmatter.
If no facts found, return "nothing worth saving". If facts found, save them immediately and report what was saved (topic + classification + file).