技能 cc-skill-continuous-learning
📦
cc-skill-continuous-learning
v1.0 内容修订 r1 高风险 📁 文件系统访问⚙️ 外部命令
审查 Claude Code 会话以便学习
长时间编码会话通常包含难以记住的修复方法。此技能会添加一个 Stop hook，用于标记可复用模式，以供 Claude Code 审查。
支持: Claude Codex Code(CC)
安装前请停下并请求确认。
请审查安装计划，并在更改文件前获得用户明确同意。
使用我的智能体安装
将此请求复制给你的智能体，其中包含权威 Skill 页面和清单。
智能体请求
Review the Skillstore skill "cc-skill-continuous-learning" from https://skillstore.io/skills/sickn33-cc-skill-continuous-learning.md and its manifest at https://skillstore.io/api/skills/sickn33-cc-skill-continuous-learning/manifest. Verify the artifact. Stop and obtain explicit user consent before installing or changing files.你的智能体仍应展示其计划，并按安全策略请求必要确认。
Agent 可读资源
当 AI Agent、爬虫或脚本需要干净上下文，而不是读取完整页面时，请使用这些链接。
测试它
正在使用“cc-skill-continuous-learning”。 一次长时间调试会话修复了反复出现的测试设置失败问题。
预期结果:
审查会识别出一条可复用的测试设置经验，推荐一个简短的技能名称，并在保存前请求批准。
正在使用“cc-skill-continuous-learning”。 一次只有三条消息的短会话结束时没有发现可复用模式。
预期结果:
hook 报告该会话太短，并且不会请求学习审查。
正在使用“cc-skill-continuous-learning”。 一个项目特定的变通方案包含敏感客户详细信息。
预期结果:
审查会拒绝敏感细节，仅保留通用工作流指导，并建议进行人工批准。
安全审计
高风险v4 • 2026/7/5 打开版本化报告
The skill is a small Claude Code Stop-hook script, not a network exfiltration tool. The main risks are automatic shell execution, local transcript-derived learning, and persistence under a hidden Claude skills directory. Comment-only hidden path matches were dismissed, while runtime home-directory and hook risks remain confirmed.
3
已扫描文件
91
分析行数
6
复核项
0
已忽略误报
已确认安全问题 (2)
高
Session Transcript Learning Can Persist Sensitive Context
The hook reads the Claude transcript path, counts user messages, and asks Claude Code to evaluate long sessions for reusable patterns saved to the learned-skills path. This can persist project-specific or sensitive lessons across future sessions if users enable it without review.
Lines 42-50 use the Claude transcript path and count messages, while lines 58-60 request evaluation and name the save location. The script does not write transcript contents itself, so confidence is high but not absolute.
中
Automatic Stop Hook Execution Requires Explicit Consent
The script is designed to run from a Claude Code Stop hook after sessions end. Users need clear opt-in because the hook runs local commands and can create persistent learning state after normal coding sessions.
The file header and hook example clearly show Stop-hook usage. The hook is manually configured, so this is a consent and transparency risk rather than covert installation.
能力复核项 (6)
这些是真实的本地能力，对此技能可能是预期行为，因此需要复核，但不会按已确认恶意行为计分。
高
Hidden file in home directory
"learned_skills_path": "~/.claude/skills/learned/",
The config sets learned_skills_path to ~/.claude/skills/learned, a hidden Claude skills directory. That location can persist generated behavior across future Claude Code sessions.
高
Hidden file in home directory
LEARNED_SKILLS_PATH=$(jq -r '.learned_skills_path // "~/.claude/skills/learned/"' "$CONFIG_FILE" | s
The script loads a learned-skills path that defaults to ~/.claude/skills/learned and later creates that directory. This can establish persistent skill state under the hidden Claude home.
高
Hidden file access
LEARNED_SKILLS_PATH="${HOME}/.claude/skills/learned"
The script initializes LEARNED_SKILLS_PATH to ${HOME}/.claude/skills/learned. It later creates that directory, which persists data under a hidden Claude skills location.
高
Hidden file access
LEARNED_SKILLS_PATH=$(jq -r '.learned_skills_path // "~/.claude/skills/learned/"' "$CONFIG_FILE" | s
The config-derived path keeps the default hidden Claude skills directory and is used for local persistence. This is a real hidden home-directory access pattern.
中
Hidden file access
"learned_skills_path": "~/.claude/skills/learned/",
The configured output path targets a hidden directory in the user home. This is intentional persistence and should be reviewed before enabling the hook.
中
Unix shell invocation
#!/bin/bash
The skill ships a Bash script intended to run as a Claude Code Stop hook. Automatic local command execution is a real risk factor even when the current commands are simple.
风险因素
📁 文件系统访问 (12)
审计者: codex 查看审计历史 →
分享与引用此报告
分享版本化评估报告、中性徽章、嵌入卡片和引用。Skillstore 只报告证据，不判断该 Skill 是否安全。
复制报告链接
https://skillstore.io/skills/sickn33-cc-skill-continuous-learning/audits/4?utm_source=security_passport&utm_medium=share&utm_campaign=versioned_reportMarkdown 徽章
[](https://skillstore.io/skills/sickn33-cc-skill-continuous-learning?utm_source=security_passport_badge)HTML 徽章
<a href="https://skillstore.io/skills/sickn33-cc-skill-continuous-learning?utm_source=security_passport_badge"><img src="https://skillstore.io/badges/skills/sickn33-cc-skill-continuous-learning/security.svg" alt="Skillstore security assessment" loading="lazy"></a>嵌入卡片
<iframe src="https://skillstore.io/embed/skills/sickn33-cc-skill-continuous-learning.html" title="Skillstore Security Assessment" sandbox="allow-popups allow-popups-to-escape-sandbox" loading="lazy" referrerpolicy="no-referrer" width="420" height="180"></iframe>学术引用 (APA · BibTeX · CFF)
APA 引用
affaan-m. (2026). cc-skill-continuous-learning security audit report (audit version 4) [Author version 1.0]. Skillstore. https://skillstore.io/skills/sickn33-cc-skill-continuous-learning/audits/4BibTeX 引用
@techreport{affaan-m-sickn33-cc-skill-continuous-learning-2026,
author = {affaan-m},
title = {cc-skill-continuous-learning security audit report (audit version 4)},
institution = {Skillstore},
year = {2026},
number = {4},
url = {https://skillstore.io/skills/sickn33-cc-skill-continuous-learning/audits/4},
note = {Author version 1.0}
}CITATION.cff
cff-version: 1.2.0
message: "If you use this Skill, cite its author and this versioned security audit report."
title: "cc-skill-continuous-learning security audit report (audit version 4)"
version: "1.0"
type: report
authors:
- name: "affaan-m"
date-released: "2026-07-05"
url: "https://skillstore.io/skills/sickn33-cc-skill-continuous-learning/audits/4"
identifiers:
- type: other
value: "skillstore:sickn33-cc-skill-continuous-learning:audit:4"
description: "Skillstore immutable audit report identifier"
Skillstore 评分
评分依据 证据置信度: 中 55
架构
85
可维护性
87
内容
69
社区
83
规范符合性
你能构建什么
捕获调试经验
审查长时间 Claude Code 会话，以发现可重复的修复方法和工作流模式。
标准化反复出现的修复方法
识别可转化为经审查的团队指导或可复用本地技能的模式。
测试学习工作流
评估 Stop hooks 如何在不增加每条消息延迟的情况下支持记忆能力。
试试这些提示
审查我上一次会话
Use the continuous learning hook output to identify one reusable debugging pattern. Show the lesson and why it is worth saving.
起草一个已学习技能
Review the session summary and draft a concise learned skill for a repeated workflow. Keep sensitive project details out.
调整检测设置
Inspect my continuous learning configuration and suggest safer threshold, pattern, and output path choices for my workflow.
启用前审计
Evaluate this hook setup for privacy, persistence, and shell execution risks. Recommend changes before I enable it across projects.
最佳实践
- 在将生成的学习内容保存到共享技能目录之前先进行审查。
- 当会话可能包含私密细节时，请使用项目本地输出路径。
- 将最小会话长度设置得足够高，以避免捕获噪声内容。
避免
- 在审查其文件系统行为之前，不要全局启用该 hook。
- 不要将原始转录记录摘录保存为可复用技能。
- 不要假设每一次长会话都包含有用模式。
常见问题
此技能会自动创建完整技能吗？
不会。该脚本会创建 learned-skills 目录，并向 Claude Code 发出审查较长会话的信号。
它会读取哪些数据？
它会从 Claude Code 读取转录记录路径，并统计该转录记录中的用户消息数量。
它需要额外工具吗？
是的。该脚本需要 Bash、jq、grep 和 Claude Code hook 支持。
我可以更改输出位置吗？
可以。在启用该 hook 之前，请更改 config.json 中的 learned_skills_path。
它适合用于私有项目吗？
请谨慎使用。手动审查输出，并避免保存敏感项目细节。
Codex 可以使用它吗？
Codex 可以检查这些文件，但自动 Stop hook 是为 Claude Code 设计的。
开发者详情
作者
affaan-m许可证
MIT
作者版本
v1.0
Skillstore 修订
r1
版本提示
作者声明的版本不符合 SemVer 规范。
仓库
https://github.com/sickn33/antigravity-awesome-skills/tree/main/skills/cc-skill-continuous-learning引用
6425ec35f4ac137735cff58cd7877843bba23e3b
维护新鲜度
2026/7/18
使用量
6 次下载 · 162 次浏览
文件结构