← 返回 Skills 市场
Session Health Monitor
作者
assistantheinrich-prog
· GitHub ↗
· v1.1.0
557
总下载
0
收藏
0
当前安装
3
版本数
在 OpenClaw 中安装
/install session-health-monitor
功能描述
Context window health monitoring for OpenClaw agents — threshold warnings via Telegram, pre-compaction snapshots, and memory rotation.
安全使用建议
This skill appears to implement local context monitoring and snapshotting, but before installing you should: (1) review the included scripts yourself, (2) back up ~/.claude/settings.local.json if you plan to run setup-statusline.sh (it will overwrite/add a statusLine entry), (3) be aware the skill does not itself send Telegram messages or require a Telegram token — it only instructs the agent to append footers to messages sent via whatever Telegram integration you already have, and (4) run the installer in a sandbox or with explicit consent to avoid surprising persistent changes. If you do not use Claude Code or do not want the skill to modify ~/.claude, avoid running setup-statusline.sh and use the scripts manually instead.
功能分析
Type: OpenClaw Skill
Name: session-health-monitor
Version: 1.1.0
The skill bundle is designed for OpenClaw agent context window health monitoring and memory management. All scripts and instructions align with this stated purpose, focusing on local file operations within designated memory directories (`~/.openclaw/workspace/memory`, `~/.claude/memory`, `/tmp`). There is no evidence of data exfiltration, unauthorized remote execution, persistence mechanisms, or malicious prompt injection attempts against the agent. The `SKILL.md` instructions guide the agent to extract 'key facts' about its own work and save them to local memory files, and to append status footers to Telegram messages, all of which are benign actions for internal agent management.
能力评估
Purpose & Capability
The scripts and instructions align with the stated goal of monitoring context usage, taking pre-compaction snapshots, rotating memory files, and exposing a statusline. However the SKILL.md mentions 'warnings via Telegram' but the bundled scripts do not implement Telegram network calls or request a Telegram token — the skill expects the agent's existing Telegram messaging tool to be used, which is reasonable but not explicitly documented as a dependency.
Instruction Scope
Runtime instructions ask the agent to append context-footers to outgoing Telegram messages and to modify agent heartbeat loops. The provided setup script (setup-statusline.sh) modifies ~/.claude/settings.local.json to add a statusLine entry and copies a statusline script into ~/.claude — this changes another application's configuration outside the skill's own files, which is scope creep and could surprise users if done without explicit consent.
Install Mechanism
No remote install or downloads are performed (instruction-only with local scripts). The only external dependency is jq for the installer/statusline patches. No network fetches or archives are used, so installation risk is low, but users should still inspect/prompt before running setup-statusline.sh since it writes into home directories.
Credentials
The skill uses sensible, optional environment variables (MEMORY_DIR, KEEP_DAYS, HEALTH_GREEN_MAX, etc.) and local filesystem paths (~/.openclaw, ~/.claude, /tmp). It does not request tokens or secrets in the registry metadata. The use of user home directories for memory and state is proportionate to the purpose.
Persistence & Privilege
Although always:false and no autonomous 'always-on' flag is set, the installer writes a persistent statusline script into ~/.claude and patches settings.local.json, creating persistent behavior across restarts of the Claude Code app. Modifying another tool's settings is a notable privilege and should be done only with explicit user consent and backup.
如何使用
- 确保已安装 OpenClaw（本地或 Docker 部署）
- 在对话框中输入安装命令：
/install session-health-monitor - 安装完成后，直接呼叫该 Skill 的名称或使用
/session-health-monitor触发 - 根据 Skill 的参数说明提供必要输入，即可获得结构化输出
版本历史
v1.1.0
**Added Telegram threshold warnings, clarified OpenClaw agent integration, and updated setup instructions.**
- Agents now warn about high context usage directly in Telegram messages via contextual footers at YELLOW/RED thresholds.
- Setup instructions updated for OpenClaw: add shared skill references and message footer logic.
- Statusline integration and Claude Code–specific content removed; focus is on agent loop, Telegram workflow, and OpenClaw support.
- Troubleshooting updated for Telegram footers and shared skill references.
- No script/code changes; documentation only.
v1.0.1
Add shared skill-session-health.md for OpenClaw agent integration
v1.0.0
Initial release: context window statusline, compaction detection, pre-compaction snapshots, memory rotation
元数据
常见问题
Session Health Monitor 是什么？
Context window health monitoring for OpenClaw agents — threshold warnings via Telegram, pre-compaction snapshots, and memory rotation. 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 557 次。
如何安装 Session Health Monitor？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install session-health-monitor」即可一键安装，无需额外配置。
Session Health Monitor 是免费的吗？
是的，Session Health Monitor 完全免费（开源免费），可自由下载、安装和使用。
Session Health Monitor 支持哪些平台？
Session Health Monitor 跨平台运行，可在任意部署了 OpenClaw / Claude Code 的环境中使用（cross-platform）。
谁开发了 Session Health Monitor？
由 assistantheinrich-prog（@assistantheinrich-prog）开发并维护，当前版本 v1.1.0。
推荐 Skills