ClawHub
Back to skill

Security audit

Content Rewriter

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed cloud-backed content rewriting tool; the main risk is that selected text files are sent to Evolink for processing.

Install only if you are comfortable sending chosen drafts to api.evolink.ai. Do not use it on secrets, regulated data, private business documents, or customer information unless Evolink's terms fit your needs, and prefer explicit approval before an agent uses it for routine editing requests.

SkillSpector

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (8)

Lp3

Medium
Category
MCP Least Privilege
Confidence
90% confidence
Finding
The skill advertises executable behavior that uses environment variables, shell commands, and file read/write via `scripts/rewriter.sh`, but it does not declare permissions explicitly. This can undermine platform trust and informed consent because users and policy layers may not realize the skill will access local files, invoke shell tooling, and transmit content off-host using an API key.

Vague Triggers

Medium
Confidence
75% confidence
Finding
The trigger phrases are very broad, such as common requests to rewrite content or change tone, which increases the chance the skill is invoked in ordinary conversations without the user specifically intending to use this third-party integration. Because the skill sends user content to an external API, accidental activation could cause unintended data disclosure.

Natural-Language Policy Violations

Medium
Confidence
84% confidence
Finding
Documenting `wechat` as producing Chinese output by default can cause the skill to transform user content into a different language purely based on platform selection, without explicit user consent to translation. This is primarily a user-intent and output-integrity issue rather than a direct security compromise, but it can lead to unexpected disclosure or misuse if users assume only formatting changes will occur.

Missing User Warnings

Medium
Confidence
86% confidence
Finding
The installer automatically copies files into a detected workspace and updates lock metadata without prompting the user or requiring an explicit destination confirmation. While this is common in package installers, it is still a real security concern because workspace detection is influenced by the current directory and environment variables, so a user can unintentionally modify an unexpected filesystem location.

Missing User Warnings

Medium
Confidence
97% confidence
Finding
The script reads arbitrary local file contents and sends them to the external EvoLink API for rewriting, scoring, or translation, but it does not present a clear user-facing warning or confirmation at the point of transmission. This creates a real privacy and data-handling risk because users may process sensitive documents without realizing their contents leave the local environment.

Missing User Warnings

Medium
Confidence
96% confidence
Finding
The script reads arbitrary local file content and sends it to a third-party API for rewriting, scoring, or translation, but it does not provide an explicit data disclosure warning or require a confirmation step at the point of transmission. In an agent/skill context, users may pass sensitive documents assuming local processing, creating a real privacy and data-handling risk.

External Transmission

Medium
Category
Data Exfiltration
Content
" "$native_prompt" "$native_content" "$native_payload" "$model"

  local response
  response=$(curl -s -X POST "$EVOLINK_API" \
    -H "Authorization: Bearer $api_key" \
    -H "Content-Type: application/json" \
    -d "@$tmp_payload")
Confidence
96% confidence
Finding
This curl call transmits user-supplied file contents to a third-party API endpoint, which is a genuine external data exfiltration path by design. In this skill's context, that is expected functionality, but it is still security-relevant because any secrets, internal documents, or personal data in the input file are exposed to an external service.

External Transmission

Medium
Category
Data Exfiltration
Content
" "$native_prompt" "$native_content" "$native_payload" "$model"

  local response
  response=$(curl -s -X POST "$EVOLINK_API" \
    -H "Authorization: Bearer $api_key" \
    -H "Content-Type: application/json" \
    -d "@$tmp_payload")
Confidence
97% confidence
Finding
This code performs an outbound HTTPS POST containing user-supplied file contents and prompts to an external API. In a security review of agent skills, undisclosed external transmission of local content is a real data exfiltration concern, especially if the input files may contain proprietary or regulated information.

Static analysis

No suspicious patterns detected.