Lobster Compact

Summarize long conversations to preserve context. Automatically triggered when context window approaches limits, or manually with /compact.

Security Scan

This instruction-only skill is coherent with its stated purpose: it summarizes long conversations and saves compacted summaries to the agent's memory files, but it will persist conversation details (including filenames, code snippets, and any secrets present) into memory storage so you should review memory handling policy before enabling it.

Purpose & Capability

The name/description (conversation compaction) matches the runtime instructions: analyze exchanges, create a structured summary, and write summaries to memory files. Nothing requested (no extra env vars or binaries) is out of scope for that purpose.

Instruction Scope

Instructions explicitly require extracting file names, code snippets, errors, and other granular context and then saving them to MEMORY.md and memory/YYYY-MM-DD.md. That is consistent with a compaction skill but means the skill will persist potentially sensitive data (API keys, passwords, private code) unless the agent sanitizes or filters content first. The instructions do not include explicit redaction or sanitization steps.

Install Mechanism

No install step or code files — instruction-only. Lowest risk from installation because nothing is downloaded or written beyond the normal memory writes described in the instructions.

Credentials

No environment variables, credentials, or external config paths are requested. The allowed tools (read/write/edit/memory_search/memory_get) are aligned with the stated memory-writing behavior.

Persistence & Privilege

The skill instructs writing to persistent memory files (MEMORY.md and dated memory logs), which is expected for a compaction tool. always:false and no modifications to other skills are declared. Still, this persistence means summaries (including sensitive content) will be stored long-term — verify memory access controls and retention policies.

Assessment

This skill appears to do what it says: summarize conversations and save those summaries to your agent's memory. Before enabling it, decide if you want full conversation details (file names, code snippets, error traces) stored persistently. Ask or configure the agent to redact secrets (API keys, passwords, private keys) before saving, confirm where memory files are stored, who can access them, and how long they are retained. If you prefer tighter control, use /compact manually rather than allowing autonomous invocation, and periodically review or prune MEMORY.md and dated memory logs.

Compact — 长会话压缩

Inputs

• $args: Optional focus area for the summary

Goal

Create a detailed summary of the conversation so far that preserves all critical context for continuing work.

Steps

1. Analyze the Conversation

Chronologically analyze each exchange:

• User's explicit requests and intents
• Approach taken to address requests
• Key decisions, technical concepts, code patterns
• Specific details: file names, code snippets, function signatures
• Errors encountered and how they were fixed
• User feedback and corrections

2. Generate Summary

Write a summary with these sections:

# 会话摘要

## 1. 主要请求和意图
[用户的核心需求]

## 2. 关键技术概念
[涉及的技术、框架、工具]

## 3. 文件和代码
[涉及的文件和关键代码片段]

## 4. 错误和修复
[遇到的问题及解决方案]

## 5. 用户反馈
[用户给出的指导、纠正、偏好]

## 6. 待办任务
[尚未完成的任务]

## 7. 当前工作
[最近在做什么，精确到文件和代码]

## 8. 下一步
[紧接着应该做什么]

3. Update Memory

• Save critical long-term info to MEMORY.md
• Save today's log to memory/YYYY-MM-DD.md
• Use memory_search to avoid duplicates

Success criteria: Summary captures all critical context; memory files updated.