← 返回 Skills 市场

Self Improver Lite

Name: Self Improver Lite
Author: utromaya-code

作者 utromaya-code · GitHub ↗ · v1.0.1 · MIT-0

cross-platform ⚠ suspicious

171

总下载

当前安装

版本数

在 OpenClaw 中安装

/install agent-self-improver

功能描述

Runs lightweight self-audits of OpenClaw behavior, finds repeated failures, proposes safe config/process improvements, and tracks what changed. Use after inc...

使用说明 (SKILL.md)

Self Improver Lite

Goal

Continuously improve reliability without risky autonomous changes.

Trigger Cases

Bot is silent or unstable
Same error appears repeatedly (rate_limit, orphaned, timeout, sendMessage failed)
After major config updates
Scheduled weekly health review

Audit Steps

Collect evidence:

systemctl is-active openclaw-gateway ollama
journalctl -u openclaw-gateway -n 120 --no-pager

Group failures by pattern and count.
Identify top 1-3 root causes.
Propose minimal, reversible fixes.
Apply only low-risk fixes automatically.
Record summary and next actions.

Auto-Allowed Changes

Session cleanup when orphaned user message loops
Gateway restart for stuck polling
Context window/maxTokens tuning within known-safe ranges
Fallback chain reordering (no key changes)

Requires Explicit User Approval

New API provider onboarding
Social network account actions
Financial/crypto operations
Credential rotation or deletion
Any destructive file cleanup beyond sessions/log rotation

Weekly Review Template

## Weekly Self-Audit
- Period: \x3Cdates>
- Uptime notes: \x3Cshort>

### Top Errors
1. \x3Cerror> — \x3Ccount>
2. \x3Cerror> — \x3Ccount>

### Changes Applied
- \x3Cchange 1>
- \x3Cchange 2>

### Measured Impact
- Response latency: \x3Cbefore -> after>
- Failed runs: \x3Cbefore -> after>

### Next Improvements
- \x3Csingle highest-impact next step>

Guardrails

Never expose secrets in reports.
Always keep a rollback path (backup old config before edits).
Prefer one change per cycle to keep causality clear.

安全使用建议

This skill is broadly consistent with its stated purpose, but it leaves important details unspecified. Before installing or enabling it: (1) require explicit, per-change user approval or a strict policy that constrains automatic actions; (2) require a dry-run mode and human review for any config edits; (3) define concrete safe numeric ranges and exact commands for 'context window/maxTokens' tuning and other automated changes; (4) ensure explicit redaction rules and test that reports never leak secrets; (5) run it in a staging environment first and verify rollback procedures; and (6) verify how the agent will obtain necessary system privileges (avoid granting blanket root access). If the author can supply precise allowed-change lists, safe ranges, and an approval workflow, my assessment would move toward benign.

功能分析

Type: OpenClaw Skill Name: agent-self-improver Version: 1.0.1 The skill bundle grants the agent high-privilege system access to perform maintenance tasks, including executing 'systemctl' and 'journalctl' commands and automatically modifying configuration files in SKILL.md. While these capabilities are aligned with the stated purpose of 'self-improvement' and reliability auditing, the broad shell access and potential for automated system changes represent a significant security risk without explicit evidence of malicious intent.

能力评估

✓ Purpose & Capability

Name/description match the instructions: the SKILL.md focuses on collecting service status and journal logs, grouping failures, proposing fixes, and applying low-risk remedies such as session cleanup, gateway restart, and config tuning — all coherent for a self-auditing/maintenance skill.

⚠ Instruction Scope

Instructions explicitly run system commands (systemctl, journalctl) and permit automated changes (session cleanup, restarting services, config tuning, fallback reordering). The document does not specify exact commands, safe numeric ranges, or precise criteria for when automatic changes are allowed, which gives an agent broad discretion to modify system state.

✓ Install Mechanism

Instruction-only skill with no install spec or code files; nothing is written to disk by the skill itself, which is low risk from an install perspective.

ℹ Credentials

The skill requests no credentials or environment variables, which is appropriate; however, the actions described (reading systemd status and journals, restarting services, editing configuration) imply the need for elevated system privileges (sudo/root) that are not declared. Confirm how the agent will obtain/require those privileges and whether access will be limited.

⚠ Persistence & Privilege

always:false is good, but the skill permits automatic application of changes and the platform allows autonomous invocation by default. Combined with vague guardrails, that raises the risk that the agent could autonomously make system changes without clear operator approval or strict limits.

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 agent-self-improver
安装完成后，直接呼叫该 Skill 的名称或使用 /agent-self-improver 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

Version 1.0.1 of agent-self-improver - No file changes detected in this version. - Overall behavior, features, and documentation remain the same.

v1.0.0

Self-improver-lite 1.0.0 – Initial Release - Introduces a lightweight self-auditing process for OpenClaw to boost reliability. - Automatically detects repeated failures and proposes safe, reversible configuration or process improvements. - Allows for auto-application of low-risk changes such as session cleanup, gateway restarts, and safe context tuning. - Requires explicit user approval for sensitive or high-risk actions, including credential and financial operations. - Provides a weekly review template and enforces key guardrails to ensure safety and traceability of changes.

元数据

Slug agent-self-improver

版本 1.0.1

许可证 MIT-0

累计安装 0

当前安装数 0

历史版本数 2

常见问题