← 返回 Skills 市场
Strategic Compact
作者
Deonte Cooper
· GitHub ↗
· v1.0.1
· MIT-0
136
总下载
0
收藏
3
当前安装
2
版本数
在 OpenClaw 中安装
/install strategic-compact
功能描述
Manual context compaction at logical workflow boundaries rather than arbitrary auto-compaction. Use after completing exploration phases, before switching maj...
安全使用建议
This skill appears coherent and low-risk for its stated purpose. Before installing or integrating it: 1) inspect the included scripts (already small and readable) and confirm you are comfortable with a small counter file being created under /tmp; 2) if you run it as a pre-tool hook or cron job, be aware it will write to /tmp frequently—ensure that behavior is acceptable and that you set COMPACT_THRESHOLD/CLAUDE_SESSION_ID appropriately; 3) consider setting a stable CLAUDE_SESSION_ID per session to avoid accidental counter sharing via the 'default' fallback; and 4) if you share a multi-user system, remember files in /tmp can be enumerated—use appropriate isolation or a different directory if necessary.
功能分析
Type: OpenClaw Skill
Name: strategic-compact
Version: 1.0.1
The 'strategic-compact' skill is designed to help manage AI context limits by suggesting manual compaction at logical workflow boundaries. The included bash script (scripts/suggest-compact.sh) simply tracks tool call counts using a temporary file in /tmp and outputs suggestions to stderr, with no evidence of data exfiltration, malicious execution, or unauthorized access.
能力评估
Purpose & Capability
The name/description (strategic compaction suggestions) aligns with the provided SKILL.md, references, and the suggest-compact.sh script. All requested/used resources (optional COMPACT_THRESHOLD, CLAUDE_SESSION_ID, writing a counter file) are proportional to the stated goal.
Instruction Scope
SKILL.md stays on-topic: it explains when/how to run the script, how to save files before compacting, and integration options (manual, heartbeat, pre-tool hook, cron). The instructions direct the agent/user to read/write local files and to run the included shell script; these actions are consistent with the purpose. Note: it suggests placing the script into hooks and cron jobs which will cause repeated local executions and writes to /tmp (expected for this use case).
Install Mechanism
There is no install spec and the skill is instruction-only with a small included script. No third-party downloads or package installs are performed—low install risk.
Credentials
No required environment variables are declared. The script optionally reads COMPACT_THRESHOLD and CLAUDE_SESSION_ID, which are reasonable for configuring thresholds and session tracking. Minor caveat: the script falls back to PPID and then the literal string 'default' for SESSION_ID; using 'default' means different sessions could share /tmp/claude-tool-count-default if the env isn't set, which may be surprising or confusing but not malicious.
Persistence & Privilege
The skill does not request always:true and does not modify other skills or system-wide settings. It writes a small counter file under /tmp for its own bookkeeping, which is ephemeral and scoped to the local system/user.
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 strategic-compact - 安装完成后，直接呼叫该 Skill 的名称或使用
/strategic-compact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
v1.0.1
Fix: replace bare code blocks with ```text for consistent rendering
v1.0.0
Initial release. Manual context compaction with suggest-compact.sh script. Adapted from everything-claude-code by @affaan-m (MIT)
元数据
常见问题
Strategic Compact 是什么？
Manual context compaction at logical workflow boundaries rather than arbitrary auto-compaction. Use after completing exploration phases, before switching maj... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 136 次。
如何安装 Strategic Compact？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install strategic-compact」即可一键安装，无需额外配置。
Strategic Compact 是免费的吗？
是的，Strategic Compact 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
Strategic Compact 支持哪些平台？
Strategic Compact 跨平台运行，可在任意部署了 OpenClaw / Claude Code 的环境中使用（darwin, linux, win32）。
谁开发了 Strategic Compact？
由 Deonte Cooper（@djc00p）开发并维护，当前版本 v1.0.1。
推荐 Skills