← 返回 Skills 市场

Low-Spec Optimizer

Name: Low-Spec Optimizer
Author: lunaviva211-sketch

作者 lunaviva211-sketch · GitHub ↗ · v1.0.0 · MIT-0

cross-platform ⚠ suspicious

258

总下载

当前安装

版本数

在 OpenClaw 中安装

/install low-spec-optimizer

功能描述

Optimize OpenClaw performance on machines with limited resources (4GB RAM or less, old CPUs). Use when (1) user mentions slow performance, high RAM usage, or...

使用说明 (SKILL.md)

Low-Spec Optimizer

Help OpenClaw run smoothly on machines with limited resources.

Quick Start

Check current resources:

bash \x3Cskill-dir>/scripts/check_resources.sh

Parse output with jq:

bash \x3Cskill-dir>/scripts/check_resources.sh | jq '.alert, .ram.pct, .cpu.pct'

Workflow

1. Check resources before heavy operations

Before spawning subagents, opening browser, or running intensive tasks:

bash \x3Cskill-dir>/scripts/check_resources.sh

Decision matrix based on alert field:

OK (\x3C 60% RAM): Proceed normally
ELEVATED (60-75%): Limit to 1 subagent, close browser first
WARNING (75-90%): Run cleanup, avoid spawning, warn user
CRITICAL (> 90%): Emergency cleanup, ask user before proceeding

2. Cleanup when needed

# Dry run first
bash \x3Cskill-dir>/scripts/cleanup_sessions.sh --dry-run

# Execute cleanup
bash \x3Cskill-dir>/scripts/cleanup_sessions.sh

# Aggressive (includes npm/pip/journal)
bash \x3Cskill-dir>/scripts/cleanup_sessions.sh --aggressive

3. Apply config recommendations

For machines with ≤4GB RAM, suggest OpenClaw config changes. See references/config-guide.md for full details.

Key recommendations:

Use free/lightweight models (hunter-alpha, GLM-4.5-air)
Set thinking to "off" unless needed
Limit concurrent subagents to 2 max
Always close browser after use
Prefer web_fetch over browser automation
Use mode: "run" for subagents (auto-cleanup)

4. Proactive monitoring (heartbeat)

During heartbeats on low-spec machines:

bash \x3Cskill-dir>/scripts/check_resources.sh | jq -r '.alert'

If alert is WARNING or CRITICAL, notify the user with specifics:

Which process is consuming most memory
How much is free
Recommended action (cleanup, close apps, restart)

Rules for constrained environments

One browser at a time — never parallel browser sessions
Close everything after use — browser, subagent sessions, temp files
Check before opening — always check RAM before heavy operations
Prefer lightweight alternatives — web_fetch > browser, mode: "run" > mode: "session"
Batch operations — combine tasks instead of spawning separate sessions
Use cron over heartbeat — for precise scheduling, use cron with isolated sessions

安全使用建议

Review the scripts before installing. Specific concerns: (1) cleanup_sessions.sh uses hard-coded /home/nvi paths — change these to $HOME or a configurable path before running, or you may delete another user's data; (2) the aggressive mode performs npm/pip cache purges and journalctl vacuum without interactive confirmation — run with --dry-run first and inspect output; (3) back up important data and test on a non-production machine; (4) prefer replacing destructive rm -rf calls with safer operations (move to a temp quarantine folder, require explicit confirmation, or restrict to paths known to belong to the agent). If the author provides a configurable path option (or the script uses $HOME) and adds an explicit interactive confirmation for destructive actions, the skill would be much closer to 'benign.'

功能分析

Type: OpenClaw Skill Name: low-spec-optimizer Version: 1.0.0 The 'low-spec-optimizer' skill is designed to monitor system resources and perform routine maintenance to improve performance on resource-constrained machines. The included scripts, 'check_resources.sh' and 'cleanup_sessions.sh', use standard Linux utilities to report on RAM/CPU usage and safely clear temporary caches (e.g., Playwright, npm, pip) and stale session data. The instructions in 'SKILL.md' and 'config-guide.md' provide legitimate operational guidance for managing OpenClaw in low-RAM environments without any evidence of malicious intent, data exfiltration, or unauthorized persistence.

能力评估

⚠ Purpose & Capability

The skill claims to be a general optimizer for low-resource machines, but the cleanup script targets hard-coded paths under /home/nvi (.openclaw/sessions, .openclaw/trash, .cache/ms-playwright). A general-purpose optimizer should use $HOME or configurable paths; hard-coded user paths indicate the code was written for a specific environment and is not proportionate to the generic description.

⚠ Instruction Scope

SKILL.md instructs running the included scripts (including an 'aggressive' cleanup). The scripts will rm -rf directories and run system-wide cleanup commands (npm cache clean, pip3 cache purge, journalctl vacuum) without prompting for confirmation unless the caller uses --dry-run. SKILL.md says to 'ask user before proceeding' for CRITICAL, but the provided scripts do not implement interactive confirmation, creating a mismatch between documentation and actions.

✓ Install Mechanism

This is an instruction-only skill with local shell scripts and no install spec or external downloads. No network retrieval or third-party package installs are performed by the skill itself, so the install mechanism risk is low.

⚠ Credentials

The skill declares no required env vars, but the scripts assume specific filesystem locations (/home/nvi/...), existence of system tools (journalctl, npm, pip3), and a particular OpenClaw layout. Requesting or operating on resources tied to a specific user without exposing that dependency in metadata is disproportionate and surprising to users on other systems.

ℹ Persistence & Privilege

The skill does not request always:true and does not modify other skills or global agent configs. Autonomous invocation is allowed by default (not flagged alone). However, because the skill can be invoked and will perform deletions on the local filesystem, its ability to run autonomously raises potential for unwanted destructive actions if invoked without user supervision.

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 low-spec-optimizer
安装完成后，直接呼叫该 Skill 的名称或使用 /low-spec-optimizer 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

v1.0.0

Initial release: resource monitoring, session cleanup, config recommendations for machines with limited RAM/CPU

元数据

Slug low-spec-optimizer

版本 1.0.0

许可证 MIT-0

累计安装 1

当前安装数 1

历史版本数 1

常见问题