← 返回 Skills 市场
115
总下载
0
收藏
0
当前安装
1
版本数
在 OpenClaw 中安装
/install session-tracker
功能描述
ALWAYS use this skill for ANY multi-step task, complex task, or task that could be interrupted. YOU MUST use session-tracker whenever working on anything wit...
安全使用建议
This skill's SKILL.md describes a CLI and background monitor that write checkpoint files and scan project directories, but the package includes no binary or install steps — that mismatch is the main red flag. Before installing or using this skill: 1) Ask the publisher for the actual implementation/binary or an install script and inspect it; 2) Verify exactly what path will be used for SESSION_DIR and whether it will touch any directories with sensitive data (especially the 'skills' directory); 3) If you proceed, run it in an isolated project or sandbox first so you can review all created files (state.json, worklog.jsonl, monitor.pid, etc.); 4) Prefer skills that provide source code or documented, auditable install steps; 5) If you cannot obtain/inspect the implementation, do not grant it persistent filesystem access or run it on sensitive projects.
功能分析
Type: OpenClaw Skill
Name: session-tracker
Version: 2.1.0
The session-tracker skill is a robust utility designed to maintain state and recover progress for multi-step tasks across agent session interruptions. It implements a background monitor process (session_tracker.py) that uses filesystem scanning and 'micro-dumps' to detect activity or 'stuck' states, and a 'Gridman Outsider' pattern to alert new agents of previous crashes via worklog.md. While the SKILL.md uses forceful instructions to ensure agent compliance and the script manages background processes, all behaviors are transparently documented, aligned with the stated purpose of task persistence, and show no evidence of malicious intent, data exfiltration, or unauthorized access.
能力评估
Purpose & Capability
The skill claims to provide a session-tracking CLI and background monitor that checkpoints state to disk. However, the registry entry contains only SKILL.md (no code, no install). The README repeatedly refers to running a 'session-tracker' command and a background monitor process, which cannot exist unless the environment already supplies that binary or the package installs it. Hardcoded default paths (/home/z/my-project/...) and explicit monitoring of a 'skills' directory are plausible for a tracker, but require filesystem access and a real implementation — which is missing.
Instruction Scope
The instructions direct the agent to create and write persistent session files, run a background monitor (monitor.pid, microdump files), and actively scan multiple directories (download, upload, .session, skills). They also instruct reading session state and checking for orphaned sessions. This goes beyond a simple stateless helper: it instructs persistent disk writes and recursive filesystem monitoring, including a directory for 'skill invocations' which could reveal other agents' outputs. Because there's no code provided, it's unclear who/what should perform these actions and how.
Install Mechanism
No install specification or code files are provided, yet SKILL.md repeatedly references a CLI and a background monitor. An instruction-only skill that requires a custom binary or daemon but provides neither is incoherent and risky: it either assumes an out-of-band installation (not documented) or it will instruct the agent to create ad-hoc scripts/processes, which is unclear and potentially unsafe.
Credentials
The skill requests no environment variables or credentials, which is proportionate in that it doesn't ask for secrets. However, it requires broad filesystem access (read metadata and write persistent state under /home/z/my-project/) and will scan a 'skills' directory. That level of disk access is reasonable for a session tracker, but should be explicit and auditable — especially because it could expose other tools' outputs or sensitive files.
Persistence & Privilege
The skill's workflow creates persistent artifacts (state.json, worklog.jsonl, microdump files, monitor.pid, sentinel file) and expects a long-running background monitor. While 'always' is false, these instructions still establish on-disk persistence and a process model that survives conversation restarts. Without concrete install code or provenance, this persistence is a cautionary flag.
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-tracker - 安装完成后，直接呼叫该 Skill 的名称或使用
/session-tracker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
v2.1.0
## What's New in v2.1
| Feature | v1 | v2 | v2.1 |
|---------|----|----|----|
| Stuck detection | Micro-dump only | FS scanner + micro-dump | Same |
| Activity detection | Manual only | Auto via `os.stat` | Same |
| File reading | Not tracked | `--reading` + atime | Same |
| Heartbeat | None | `ping` command | Same |
| File renames | Not tracked | `--rename` | Same |
| TodoWrite sync | None | `sync` command | Same |
| Resume info | Steps only | Steps + worklog | Same |
| **Meta-crash detection** | None | None | **ACTIVE sentinel + `crash-detect`** |
| **Orphan auto-detection** | None | None | **`init` warns about crashed sessions** |
| **Crash marker in worklog.md** | None | None | **Any new agent sees the crash** |
| **Recovery report** | None | None | **`crash-detect` command** |
元数据
常见问题
Session Tracker 是什么？
ALWAYS use this skill for ANY multi-step task, complex task, or task that could be interrupted. YOU MUST use session-tracker whenever working on anything wit... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 115 次。
如何安装 Session Tracker？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install session-tracker」即可一键安装，无需额外配置。
Session Tracker 是免费的吗？
是的，Session Tracker 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
Session Tracker 支持哪些平台？
Session Tracker 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 Tracker？
由 DarkD（@darkd）开发并维护，当前版本 v2.1.0。
推荐 Skills