Back to skill

Security audit

WuCai

Security checks for vulnerabilities and agentic risk

Overview

This appears to be a legitimate WuCai integration, but it needs review because it asks users to paste a live API token into chat and can modify or trash private knowledge-base items without explicit confirmation safeguards.

Review this skill before installing. Use a protected secret or environment setting for WUCAI_API_TOKEN instead of pasting it into chat, rotate the token if it has already been pasted, prefer a ClawHub or pinned-version install instead of the mutable raw GitHub URL, and require explicit confirmation before allowing the agent to update notes, append diary entries, change statuses, or trash articles.

Vulnerability Patterns
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • 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
Findings (2)

T08 · Insecure Dependencies

Warning
Location
README.md:19
Finding

Installation from a Mutable Remote Source

Content
View full analysis
Remediation
View remediation
/SKILL.md ``` 2. Prefer signed, versioned release artifacts over direct installation from a development branch. 3. Publish a SHA-256 digest for every release and require the installer to verify it before loading the skill. 4. Use signed Git tags or artifact signatures and document how users or the installer can verify them. 5. Ensure the pinned skill and all referenced scripts belong to the same immutable release. 6. Retain the human-readable version number, commit hash, and checksum in the installation documentation to make the audited artifact identifiable. ]]>

T09 · Insecure Skill Coding Practices

Warning
Location
README.md:52
Finding

Documentation Encourages Disclosure of an API Token Through Chat

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Taint TrackingDirect Taint Flow, Variable-Mediated Taint Flow, Credential Exfiltration Chain
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (11)

Tainted flow: 'req' from os.getenv (line 55, credential/environment) → urllib.request.urlopen (network output)

Critical
Category
Data Flow
Confidence
90% confidence
Finding

Credentials or environment variables flow to a network sink. This is a high-confidence indicator of credential exfiltration.

Content

Scanner excerpt · scripts/wucai_api.py (reported line 58)May include surrounding context.

python
req = urllib.request.Request(full_url, data=data, headers=headers, method='POST')
        
        # 5. 执行请求,强制 15s 超时
        with urllib.request.urlopen(req, timeout=15) as response:
            res_body = response.read().decode('utf-8')
            try:
                return json.loads(res_body)

Ssd 3

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The skill explicitly tells users to paste their authentication token into the AI conversation to complete binding. Because this skill manages highly sensitive personal knowledge-base content such as highlights, annotations, and diary entries, exposing the token in chat can let the agent platform or any compromised component access private data across the user's account.

Content

No source excerpt is available for this finding.

Tp2

High
Category
MCP Tool Poisoning
Confidence
85% confidence
Finding

Mixing characters from multiple Unicode scripts in a single identifier is a common technique to create visually ambiguous tool names.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The README instructs users to send a live authentication token directly to the AI for account binding, but does not clearly warn that the token is a sensitive credential with account-level access. In an agent ecosystem, pasting secrets into chat increases the chance of unintended storage, logging, prompt leakage, or reuse by other tools/plugins, making credential compromise a realistic risk.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
86% confidence
Finding

The skill invokes a Python script that reads environment variables and performs API/network operations, but it does not declare an explicit tool scope such as allowed tools or permissions. This creates an authorization ambiguity: an agent runtime may grant broader execution or network capability than intended, making the skill more dangerous if the script is modified, replaced, or abused to access secrets like the API token.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The instruction says the AI must decide response language solely from WUCAI_REGION, forcing Chinese for cn and English for eu/us. This is a natural-language locale policy issue because it removes user choice and ties language to region rather than explicit preference.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The skill documents destructive operations such as trashing articles and changing status, but it does not require confirmation, warn about data loss, or distinguish reversible from irreversible actions. In an agent setting, this increases the risk of unintended user-data modification or deletion from ambiguous prompts, summarization mistakes, or prompt injection relayed through content.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
84% confidence
Finding

The section title and table label '错误处理与双语话术 (Error Handling)' present prescribed CN/EN phrasing, but the broader document is written as instructions for the AI primarily in Chinese and does not state that the user may choose their preferred language. This can violate a language/locale policy if the skill defaults to a specific language without explicit user opt-in.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
86% confidence
Finding

This code forwards arbitrary caller-supplied JSON to an external WuCai API endpoint without any consent, minimization, or warning mechanism in the file. In the context of a personal knowledge-base skill handling highlights, notes, clippings, and diary data, this can expose highly sensitive user content to a remote service unexpectedly.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
91% confidence
Finding

The manifest description is written only in Chinese and does not indicate that the skill supports multiple languages or that it is intentionally limited to a Chinese-speaking audience. Under the stated policy, forcing a specific language without opt-in can be a natural-language policy concern.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
78% confidence
Finding

The script reads an API token from the WUCAI_API_TOKEN environment variable and uses it for authentication, but this file does not include a user-facing notice that it accesses credentials from the environment. Comments about validation are present, but they do not disclose the behavior to the user.

Content

No source excerpt is available for this finding.