Safe Config Workflow
PassAudited by VirusTotal on May 11, 2026.
Overview
Type: OpenClaw Skill Name: safe-config-workflow-zh Version: 1.0.0 The skill bundle is designed for safe configuration management of OpenClaw, involving standard administrative tasks like checking, fixing, and verifying configuration files. All commands specified in SKILL.md and TEST-GUIDE.md (e.g., `openclaw doctor --fix`, `diff`, `openclaw gateway status`, `nano ~/.openclaw/openclaw.json`) are directly related to this stated purpose. The instructions for the AI agent emphasize safety, user confirmation, documentation lookup, and learning. There is no evidence of intentional harmful behavior such as data exfiltration, unauthorized remote execution, persistence mechanisms, or deceptive prompt injection attempts. The use of `nano` is for manual configuration repair in a troubleshooting context, not for arbitrary file manipulation or malicious code injection.
Findings (0)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
The agent could change more of the OpenClaw configuration than the user specifically requested, potentially affecting Gateway behavior, channels, sessions, or authentication settings.
The skill directs the agent to run an automatic repair command against the OpenClaw config and treats small fixes as not requiring user involvement. This is purpose-aligned, but the exact scope of automatic changes is not bounded beyond the doctor tool.
步骤 2: 检查 + 自动修复 openclaw doctor --fix ... 小问题 ... 不麻烦用户
Require explicit user approval before applying doctor --fix changes, show a clear diff before and after, and provide an easy restore path from the backup.
Authentication or session-related configuration values could be exposed in chat output, diffs, logs, or memory if they appear in the config.
The skill explicitly covers session and authentication configuration and instructs the agent to report old and new config values, but it does not include redaction rules for tokens, secrets, or auth-related values.
修改模型/会话/认证配置 ... 字段：XXX ... 原值："XXX" ... 新值："XXX"
Add rules to never display, copy, or persist secrets; redact token-like values in diffs and summaries; and ask the user before touching auth/session settings.
Private configuration details could persist across future tasks and influence later agent behavior without the user realizing it.
The skill mandates persistent learning in MEMORY.md and frames it as a lasting knowledge base, but does not define what may be stored, how long it persists, or how sensitive config values are excluded.
把学到的正确配置值记录到 MEMORY.md ... 形成知识库，永久避免
Store only nonsensitive schema lessons, not actual user config values; redact secrets; define the memory file location and retention; and get user consent before writing persistent notes.
Testing or applying config changes at the wrong time could temporarily interrupt Gateway service.
The test guide acknowledges that some config tests and restarts can temporarily break Gateway availability. This is disclosed and includes backup/restore steps, so it is a note rather than a standalone concern.
⚠️ 警告：这会导致 Gateway 暂时无法运行 ... openclaw gateway restart
Run disruptive tests only during a safe maintenance window, keep backups, and verify Gateway status after every change.