T09 · Insecure Skill Coding Practices
- Location
src/monitor.js:49- Finding
OS Command Injection in DingTalk Notification Handling
- Content
View full analysis
- Remediation
View remediation
Security audit
Security checks for vulnerabilities and agentic risk
This announcement monitor has a coherent purpose, but it can run shell commands from scraped or configured text and includes a default external DingTalk recipient, so it needs review before installation.
Review this skill carefully before installing. Replace all example notification targets with your own, run only with trusted HTTPS URLs, avoid enabling cron until notification delivery is verified, and prefer a fixed version that removes shell-based message sending and runs Chromium with sandboxing enabled.
src/monitor.js:49OS Command Injection in DingTalk Notification Handling
src/monitor.js:72Chromium Sandbox Disabled While Rendering Arbitrary Websites
configs/default.yaml:28Hardcoded DingTalk Recipient Causes Unintended Third-Party Disclosure
Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).
rm ~/.openclaw/workspace/state/pushed-ids-*.json
Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).
rm ~/.openclaw/workspace/state/pushed-ids-*.json
basic-ftp 5.2.0 is present transitively via get-uri/proxy tooling and has reported command-injection and DoS issues. Even though this skill is not obviously an FTP client from the lockfile alone, vulnerable packages in the dependency tree can become reachable when proxy/PAC/URI handling paths are exercised, especially in automation tooling like Puppeteer.
extract-zip 2.0.1 has advisories for symlink-based arbitrary write/path traversal during archive extraction. In this dependency tree it is pulled in by @puppeteer/browsers, which increases relevance because browser automation tools often download and unpack browser binaries automatically, making archive extraction a realistic code path.
Dependency has known vulnerabilities (CVEs). Using packages with unpatched security flaws exposes the environment to known exploits.
js-yaml 4.1.1 has multiple CPU-exhaustion advisories related to parsing malicious YAML structures. Because the project explicitly depends on yaml and also includes js-yaml transitively through cosmiconfig, configuration parsing is a plausible code path, so hostile YAML input could trigger denial of service if untrusted config files are processed.
tar-fs 3.0.4 has path traversal and symlink/link-following extraction issues. It is included under @puppeteer/browsers, making it meaningfully reachable in workflows that download and unpack browser artifacts; a crafted archive could write outside the intended directory or abuse predictable extraction targets.
ws 8.16.0 has multiple denial-of-service and memory-safety advisories. Since ws is a direct Puppeteer dependency used for DevTools/browser communication, it is more likely to be exercised than many other transitive packages; if exposed to malicious or intermediary-controlled websocket traffic, the skill could crash or leak memory.
The README documents a webhook notification mode that posts monitored notice content to an arbitrary external URL, but it does not include any warning about transmitting potentially sensitive monitoring results or metadata off-system. For markdown files, this falls under missing user warnings about behaviors that could affect privacy or system integrity.
The webhook example encourages sending monitored data and an Authorization bearer token to an external endpoint without any warning about secret handling, TLS validation, endpoint trust, or data exposure. In a monitoring skill that routinely forwards collected notices, this can lead to accidental credential leakage or unintended exfiltration to third-party services.
The description and author fields are written only in Chinese, with no indication that the skill is intended exclusively for Chinese-speaking users or that language selection is configurable. This can violate a language/locale policy when skills are expected to offer user choice or clearly document justified locale constraints.
The notifier builds a shell command by interpolating untrusted configuration data (config.target) and message content into execSync(...). Although double quotes in the message are escaped, shell metacharacters such as $(), backticks, and quote-breaking sequences in config.target can still lead to command injection, causing arbitrary command execution under the privileges of the monitoring process.
The report generation hard-codes the 'zh-CN' locale and Chinese-language output strings, which imposes a specific language/locale on users without any opt-in or configuration path. This matches the policy category for language or locale constraints that are not optional or justified.
The skill documentation uses Chinese throughout and does not indicate that the language is optional, user-selectable, or required for a region-specific compliance reason. This can be a language/locale policy issue when a skill effectively forces a specific language without user opt-in.
The troubleshooting section recommends removing pushed-ids-*.json, which is a destructive file deletion operation, but it does not warn users that this will erase state and may cause previously sent notices to be pushed again. In markdown documentation, destructive operations should be accompanied by an explicit warning about their impact.
The skill description repeatedly documents outbound DingTalk notifications using explicit target identifiers, but does not disclose that matched announcement data will be pushed to external messaging platforms. For a markdown skill description, this is a missing user warning about data sharing and privacy impact.
The notification parameter table documents notify as supporting only dingtalk / wechat / email, but later the same file provides a concrete webhook notification configuration example. This is an active contradiction in the skill's documentation about what notification modes the skill supports, which can mislead users about the actual behavior and scope.
The manifest description is written entirely in Chinese, which indicates a language-specific user experience but provides no opt-in, alternative language, or justification for restricting the skill to that locale. The policy requires flagging language or locale constraints unless the skill explicitly offers a choice or clearly documents a justified regional scope.
The file’s natural-language content is entirely in Chinese, including comments, task names, and sample keywords, with no indication that the skill is region-specific or that users may choose another language/locale. Under the policy for natural-language violations, this can be treated as forcing a specific language without explicit opt-in or documented justification.
This manifest contains user-facing natural-language fields entirely in Chinese, including the example title and task name, but does not state that the skill is Chinese-only or offer any language/locale choice. Under the policy, forcing a specific language without user opt-in can be a natural-language policy violation.
This manifest-style YAML lists many broad keywords such as '教育', '学校', and '培训' without describing matching logic, exclusion conditions, or negative examples. In a skill configuration, that can make invocation or filtering behavior unclear and increase the chance of unintended matches.
This YAML contains user-facing natural-language labels and comments entirely in Chinese, such as the example title and task name, with no indication that language selection is optional. Under the policy, forcing a specific language without user opt-in can be a natural-language policy violation.
This JSON manifest identifies the package as "notice-monitor" and exposes a CLI binary, but provides no information about when or how the skill should be invoked, nor any limiting conditions. For manifest files, absence of specific trigger scope or exclusion conditions can make activation semantics ambiguous in downstream tooling or documentation.
Dependency has known vulnerabilities (CVEs). Using packages with unpatched security flaws exposes the environment to known exploits.
yaml 2.8.2 is reported vulnerable to stack exhaustion on deeply nested YAML input. Because yaml is a direct dependency, parsing attacker-controlled YAML could cause process crashes, though the likely consequence is denial of service rather than code execution.