Security audit

Ai Daily News By Fang

Security checks for vulnerabilities and agentic risk

Overview

This is a disclosed AI news digest skill that saves a local report and sends it to a user-configured Feishu recipient, with setup choices the user should review.

Install only if you want generated AI news digests archived locally and sent through Feishu. Before use, replace the Feishu open_id with your intended recipient, adjust the archive path if needed, and verify that the referenced smart-web-fetch helper is trusted on your machine.

SkillSpector

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access

Findings (5)

Description-Behavior Mismatch

High

Confidence: 95% confidence
Finding: The skill metadata says it should trigger when a user asks for AI news, but the body instructs the agent to push completed reports to Feishu immediately without waiting for a request. That creates unauthorized outbound communication and turns a user-invoked summarization skill into a proactive notification mechanism, which can leak collected content or operate outside user expectations and consent.

Context-Inappropriate Capability

Medium

Confidence: 87% confidence
Finding: A news collection skill does not inherently need messaging capabilities, yet this skill requires direct Feishu delivery of the full report. That expands the skill's privilege and data flow beyond what is necessary for summarization, increasing the risk of unintended disclosure or misuse of messaging channels.

Intent-Code Divergence

High

Confidence: 96% confidence
Finding: The skill contains conflicting instructions: it claims to run when the user asks for AI updates, but later mandates immediate Feishu pushing without waiting for the user. This inconsistency is security-relevant because it can bypass normal user-trigger boundaries and justify autonomous actions the user did not request.

Missing User Warnings

Medium

Confidence: 81% confidence
Finding: The skill instructs the agent to write files into the workspace automatically, but does not warn the user that local state will be modified. Silent file creation is a meaningful side effect that can surprise users, clutter storage, or be abused as a persistence mechanism if combined with other behaviors.

Missing User Warnings

High

Confidence: 94% confidence
Finding: The skill requires immediate Feishu message pushing without a privacy notice or explicit confirmation, even though the content and recipient identifier may reveal user interests, workflow details, or generated summaries to an external service. Because this is an outbound transfer to a third-party channel, the lack of consent and transparency makes the behavior significantly more dangerous in context.

Static analysis

No suspicious patterns detected.