Claude 持久化记忆系统初始化
PassAudited by ClawScan on May 1, 2026.
Overview
This instruction-only skill appears purpose-aligned: it scans the current project and writes local Claude memory files, with the main caution that those persistent files may affect future agent behavior.
This looks like a local project-memory initializer rather than a malicious skill. Before installing or invoking it, make sure you are in the correct project directory, check the generated .claude/memory files and CLAUDE.md, and avoid committing those files if they include private project details you do not want shared.
Findings (2)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
The agent may inspect project structure and source metadata in the current working tree.
The skill directs the agent to enumerate local project directories and use project-scanning commands. This is expected for a memory-initialization skill, but users should run it only in the intended project directory.
find . -maxdepth 3 -type d ... | head -80 | sort
Invoke the skill only from the project you want indexed, and review the generated files before committing or relying on them.
Generated memory files may shape future agent behavior and may contain summarized details about the project.
The skill creates persistent memory and instruction files that future Claude sessions may use. This is the advertised purpose, but persistent project-derived context can carry forward inaccurate or unintended rules.
创建 `.claude/memory/` 目录 ... 生成完整的记忆文件和 CLAUDE.md 指令
Review .claude/memory and CLAUDE.md after generation, especially before sharing or committing them.