中文 EN
Start Review
Skill Trust Decision

openclaw-release-validation

OpenClaw 发布验证技能，用于测试 OpenClaw 网关更新。所有敏感操作（OCM 安装、GitHub API 调用、文件写入）均已在文档中明确声明，凭证隔离机制完善，无阴影功能。唯一风险点为 curl|bash 远程脚本执行，但来源为可信 GitHub 官方仓库。

Install decision first Source: ClawHub Scanned: Aug 31, 2026
Files 7
Artifacts 6
Violations 0
Findings 3

Why this conclusion was reached

1/4 dimensions flagged
Pass
Declared vs actual capability

Declared resources and inferred behavior are broadly aligned.

Block
Hidden execution and egress

1 high-risk artifacts or egress signals were extracted.

Pass
Attack chain and severe findings

There is no explicit malicious chain in the report.

Pass
Dependencies and supply chain hygiene

Dependencies are present but no obvious high-risk issue stands out.

What drove the risk score up

curl|bash 远程脚本执行 +15

SKILL.md:353 - 从 GitHub 下载并执行 OCM 安装脚本，虽为行业惯例但存在供应链风险

访问 ~/.openclaw 敏感路径 +5

需要访问 OpenClaw 配置和状态目录，但已声明且用于网关验证

GitHub API 写入权限 +5

通过 gh 命令创建 issue 和 comment，但需要用户明确审批才能发布

Most important evidence

Medium Supply Chain

OCM 安装使用 curl|bash 模式

SKILL.md:353 使用 `curl -fsSL https://github.com/openclaw/ocm/releases/latest/download/install.sh | bash` 安装 OCM，这是供应链风险模式。虽来源为可信 GitHub 仓库，但下载并执行远程脚本存在中间人攻击风险。

SKILL.md:353
建议使用预签名二进制下载替代 curl|bash，或在使用前向用户明确展示将要执行的脚本内容。
Low Sensitive Access

访问 ~/.openclaw 配置目录

技能需要访问 OpenClaw 的用户配置目录来发现网关状态和执行更新，这是网关验证的必要操作。

SKILL.md:176
已声明且必要，无需修改。
Info Priv Escalation

Docker 容器管理权限

在启用本地诊断时需要运行 Docker 容器（OTLP collector），包含端口映射和文件系统挂载。

SKILL.md:590
容器使用安全加固参数（--read-only, --cap-drop=ALL, --pids-limit）且仅绑定到 127.0.0.1，设计合理。

Declared capability vs actual capability

Shell Pass
Declared WRITE
→
Inferred WRITE
SKILL.md:353-357 (OCM安装), SKILL.md:189-196 (git ls-remote), SKILL.md:240-280 (gh workflow)
Filesystem Pass
Declared WRITE
→
Inferred WRITE
SKILL.md:441-444 (创建 worksheet), SKILL.md:607-620 (telemetry 配置)
Network Pass
Declared READ
→
Inferred READ
SKILL.md:205 (fetch scorecard), SKILL.md:304-305 (gh api), scripts/check-update.mjs:318-329 (ClawHub API)
Environment Pass
Declared READ
→
Inferred READ
scripts/check-update.mjs:181-197 (OPENCLAW_HOME, OPENCLAW_STATE_DIR)

Suspicious artifacts and egress

Critical Dangerous Command
curl -fsSL https://github.com/openclaw/ocm/releases/latest/download/install.sh | bash

SKILL.md:353

Medium External URL
https://docs.openclaw.ai/maturity/scorecard.md

SKILL.md:205

Medium External URL
http://127.0.0.1:

SKILL.md:623

Medium External URL
https://clawhub.ai

scripts/check-update.mjs:13

Medium External URL
https://clawhub.ai/user/openclaw

skill-card.md:9

Medium External URL
https://clawhub.ai/openclaw/skills/release-validation

skill-card.md:39

Dependencies and supply chain

PackageVersionSourceKnown vulnNotes
opentelemetry-collector-contrib 0.104.0 docker No 本地诊断收集器，仅在用户明确启用时使用，使用安全加固参数

File composition

7 files · 1586 lines
Markdown 6 files · 1206 linesJavaScript 1 files · 380 lines
Files of concern · 3
SKILL.md Markdown · 904 lines
OCM 安装使用 curl|bash 模式 · 访问 ~/.openclaw 配置目录 · Docker 容器管理权限 · curl -fsSL https://github.com/openclaw/ocm/releases/latest/download/install.sh | bash · https://docs.openclaw.ai/maturity/scorecard.md · http://127.0.0.1:
scripts/check-update.mjs JavaScript · 380 lines
https://clawhub.ai
skill-card.md Markdown · 60 lines
https://clawhub.ai/user/openclaw · https://clawhub.ai/openclaw/skills/release-validation
Other files · structured-report.md · tooling-feedback.md · tooling-feedback.md · validation-worksheet.md

Security positives

所有敏感操作均在 SKILL.md 中明确声明，无阴影功能
严格的敏感信息排除规则：不输出凭证、本地路径、网关名称等
用户审批机制完善：in-place 更新需要明确 `approve in-place update` 回复
候选评估报告需要用户审批才能发布到 GitHub
OCM 隔离模式提供容器化安全保障
工具反馈包（tooling-feedback）机制将工具失败与候选评估隔离
结构化报告包含严格的 schema 验证和大小限制
技能来源为可信的 openclaw 官方账户（CLAWHUB 验证）