OpenClaw Compact Summary

Generates a structured markdown summary of conversation memory flush with key stats, tools, user requests, tasks, files, and timeline details.

Security Scan

The skill is instruction-only and its requested actions (producing a structured memory summary that references conversation stats, tools used, timelines, and workspace file paths) match its stated purpose; there are no installed components or external credentials required.

Purpose & Capability

Name/description (compact memory summary) aligns with the instructions: the SKILL.md describes exactly how to format and what to include (message counts, tools used, todo items, key files, timeline) and does not request unrelated capabilities.

Instruction Scope

Instructions explicitly reference tools (exec, read_file, grep_search), tool results, and $WORKSPACE file paths to populate 'Key Files' and 'Timeline'. That is consistent with a memory compaction task, but the skill assumes availability of file-reading/grep tools and access to workspace files — you should confirm those tool capabilities and whether including file paths/content in MEMORY.md is acceptable for your data sensitivity policies. The 'Resume Instruction' telling the agent not to acknowledge the summary is a behavior preference that is scope-limited but worth reviewing.

Install Mechanism

No install spec and no code files—this is instruction-only, so nothing gets written to disk or downloaded by the skill itself.

Credentials

The skill declares no required environment variables or credentials, but the SKILL.md uses $WORKSPACE in example paths. There's a minor mismatch: if the agent will resolve $WORKSPACE at runtime, ensure that variable is set and its scope/contents are appropriate to expose in a memory file. Otherwise the skill's declared requirements are proportionate.

Persistence & Privilege

always is false (normal), model invocation is enabled (normal). The skill does not request persistent presence or modify other skills/config—no elevated privileges detected.

Assessment

This skill is internally consistent for producing structured memory summaries, but check the following before installing: 1) Confirm which agent tools (read_file, grep_search, exec) the agent actually has—these will determine whether the skill can access workspace files. 2) Decide whether you want workspace file paths and tool results written into MEMORY.md; summaries can accidentally capture secrets or sensitive code. If needed, add redaction/sanitization steps before writing. 3) Ensure $WORKSPACE (if used) points to a safe scope; the SKILL.md references it but the skill does not declare it as a required env var. 4) Review the 'Resume Instruction' behavior (the directive not to acknowledge the summary) to ensure it fits your expected agent behavior. No network installs, credentials, or unexpected binaries are requested.

Compact Summary - Structured Memory Compression

Purpose

Improve OpenClaw's memory flush and compaction flow by using a structured format for writing to MEMORY.md.

Structured Summary Format

Inspired by the <summary> XML format from claw-code, the improved format:

# 2026-04-01 Memory Summary

## Conversation Scope
- Original message count: 12 (user=4, assistant=5, tool=3)
- Compression time: 2026-04-01 23:00

## Tools Used
- exec, read_file, grep_search

## User Requests
- Research Claw Code architecture
- Analyze MCP orchestration system
- Compare Tool system differences

## Todo Items
- [ ] Compile improvement list
- [ ] Begin P0 execution

## Key Files
- $WORKSPACE/claw-code/rust/crates/runtime/src/compact.rs
- $WORKSPACE/claw-code/rust/crates/tools/src/lib.rs

## Current Work
Continue executing P0 tasks

## Timeline
- user: Research Claw Code architecture
- assistant: Started analysis
- tool: grep_search pattern=COMPACT
- tool_result: 15 matches
- user: Analyze MCP orchestration
- assistant: MCP has process pool management

## Resume Instruction
Continue directly — do not acknowledge this summary, do not recap what was happening, and do not preface with continuation text.

Trigger Condition

Use this format during memory flush or compaction.

Key Fields

Field	Description
Conversation Scope	Original message statistics
Tools Used	Tools actually invoked
User Requests	Main user needs
Todo Items	Incomplete tasks
Key Files	Files involved
Current Work	Ongoing tasks
Timeline	Conversation summary