ClawHub
Back to skill

Security audit

Weibo Publish

Security checks for vulnerabilities and agentic risk

Overview

The skill is clearly for posting to Weibo, but it combines public-account posting with existing browser sessions and required shell cleanup commands that are not safely scoped.

Review before installing. Use this only when you intend the agent to post to the logged-in Weibo account, and require explicit approval before clicking send. Avoid running the documented wildcard or placeholder cleanup commands; delete only known temporary upload files after verifying their exact paths.

A.I.G

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs

SkillSpector

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Tool MisuseTool Parameter Abuse, Chaining Abuse, Unsafe Defaults
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (9)

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
94% confidence
Finding

The cleanup step uses a shell command with a placeholder filename, which encourages direct interpolation into rm /tmp/openclaw/uploads/<filename>. If <filename> is derived from untrusted input, an agent may construct unsafe shell commands, leading to command injection or unintended file deletion; even without injection, path traversal or wildcard misuse could delete files outside the intended target.

Content

Scanner excerpt · SKILL.md (reported line 85)May include surrounding context.

bash
# Remove uploaded image from temp directory
rm /tmp/openclaw/uploads/<filename>

# Optional: Clean up browser screenshots older than 7 days
find ~/.openclaw/media/browser/ -type f -mtime +7 -delete

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
97% confidence
Finding

The example includes exec "rm /tmp/openclaw/uploads/*", a wildcard deletion command exposed through a general execution tool. Wildcard destructive commands are especially dangerous in agent settings because quoting, path resolution, or later edits can broaden scope, and the presence of an exec primitive means the skill is teaching the model to perform destructive shell actions beyond posting content.

Content

Scanner excerpt · references/workflow-examples.md (reported line 27)May include surrounding context.

browser action=close targetId=

// 7. Cleanup temporary files (if any) // exec "rm /tmp/openclaw/uploads/*" // Uncomment if files were uploaded

text

## Example 2: Image Post

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
90% confidence
Finding

Although narrower than the wildcard variant, rm /tmp/openclaw/uploads/image.jpg still demonstrates destructive shell usage via an unconstrained command channel. In agent workflows, normalizing direct rm commands can be copied to other paths or parameterized unsafely, creating a latent deletion primitive unrelated to the core browser-posting task.

Content

Scanner excerpt · references/workflow-examples.md (reported line 68)May include surrounding context.

bash
// 9. Cleanup - remove uploaded image from temp directory
rm /tmp/openclaw/uploads/image.jpg

Example 3: Multiple Images

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
96% confidence
Finding

The documented pattern rm /tmp/openclaw/uploads/<filename> encourages parameterized file deletion, which is dangerous because placeholder substitution may accept untrusted or malformed values. In an LLM-agent context this can enable path traversal, accidental deletion of arbitrary files, or unsafe command construction when the placeholder is filled dynamically.

Content

Scanner excerpt · references/workflow-examples.md (reported line 101)May include surrounding context.

md
- Use `browser action=close targetId=<id>`

5. **Cleanup Temporary Files**: Temporary files are NOT automatically cleaned up
   - Remove uploaded images: `rm /tmp/openclaw/uploads/<filename>`
   - Clean old screenshots: `find ~/.openclaw/media/browser/ -type f -mtime +7 -delete`
   - **Cleanup is a required step** - always run it after publishing

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The skill automates publishing to a public social media platform but does not explicitly warn that supplied text and images will be transmitted to Weibo and potentially exposed publicly. This creates a meaningful privacy and consent risk because users or upstream agents may invoke the skill without a clear understanding that local content is leaving the system and becoming externally visible.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The examples instruct use of shell commands to copy files into /tmp/openclaw/uploads/ and later delete them, introducing local file read/write behavior outside the stated browser-automation scope. Even if intended to support uploads, this broadens the operational surface and can lead agents to manipulate local files without clear authorization boundaries.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The workflow documents shell-based cleanup commands, including recursive-style deletion patterns and find ... -delete, which expand the skill's effective capability from browser posting into arbitrary local filesystem mutation. In an agent context, normalizing these commands as a 'required step' increases the chance that an LLM will execute destructive cleanup automatically, potentially deleting unintended files if paths or placeholders are substituted incorrectly.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The cleanup guidance includes destructive commands but provides no warning, dry-run pattern, or confirmation step, despite marking cleanup as required. In an automated agent workflow, that omission materially raises the risk of accidental deletion from path mistakes, variable substitution errors, or over-broad command generation.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
87% confidence
Finding

The document hard-codes Chinese UI text such as "分享新鲜事…" and "发送" as the expected interface labels, which presumes a specific locale. There is no note that these labels depend on language settings or alternative guidance for other locales.

Content

No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.