Save Article With Images

• Outcome: Review
• Latest audit: May 28, 2026 · 2:17 PM UTC
• Version: 1.0.1

Overview

This skill saves articles as described, but it also treats sending saved files to Feishu as a normal workflow without clear opt-in or recipient scoping.

Review before installing. Use it only if you are comfortable with article content being fetched through external services, saved persistently, and potentially sent to Feishu. Confirm the URL, output directory, and Feishu recipient each time, and avoid private or sensitive articles unless external sharing is intentional.

SkillSpector

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep
• Rogue Agent: Self-Modification, Session Persistence
• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• MCP Least Privilege: Underdeclared Capability, Wildcard Permission, Missing Permission Declaration

Findings (9)

Lp3

Medium

• Category: MCP Least Privilege
• Confidence: 95% confidence
• Finding: The skill clearly performs network access and local file writes, but those capabilities are not declared in the manifest. Undeclared capabilities reduce transparency and informed consent, making it easier for a user or host agent to invoke behavior with broader access than expected.

Description-Behavior Mismatch

Medium

• Confidence: 90% confidence
• Finding: The skill is presented as saving articles locally, but the workflow also sends content-derived files to Feishu and fetches article content through the external Jina Reader service. This creates a material disclosure gap: article contents and generated artifacts may leave the local environment without the user being clearly warned.

Context-Inappropriate Capability

Medium

• Confidence: 92% confidence
• Finding: Sending generated files to Feishu is broader than the stated purpose of local article saving and introduces exfiltration risk for potentially sensitive content. Users may reasonably expect only local storage, not automatic transmission to an external messaging platform.

Context-Inappropriate Capability

Medium

• Confidence: 89% confidence
• Finding: The WeChat flow changes directories into another tool repository and executes Python from that external location. Invoking external code outside the skill's documented scope expands the trust boundary and can introduce unreviewed behavior, including arbitrary network access or file operations.

Vague Triggers

Medium

• Confidence: 84% confidence
• Finding: Broad triggers like 'save article' and 'download article' can match common user requests and cause the skill to activate in contexts the user did not intend. Because the skill writes files, accesses networks, and may send files to Feishu, accidental invocation has meaningful side effects.

Vague Triggers

Medium

• Confidence: 86% confidence
• Finding: The triggers section lists ambiguous activation phrases without clarifying when the skill should not run. In a system with many skills or natural-language routing, this raises the risk of unintentional execution of network, file, and messaging actions.

Missing User Warnings

Medium

• Confidence: 94% confidence
• Finding: The workflow omits an upfront warning that article content and generated files may be sent to Feishu. This is dangerous because users may provide or process sensitive content under the assumption the operation is local-only, leading to unintended disclosure.

Missing User Warnings

Medium

• Confidence: 87% confidence
• Finding: The script automatically fetches remote content and writes multiple files under a fixed workspace path without prompting the user, dry-run support, or quotas. In an agent setting, this can enable silent persistence of untrusted remote data, disk consumption, and unintended overwriting/collision behavior, making the skill more dangerous because saving downloaded content is its core function.

Session Persistence

Medium

• Category: Rogue Agent

Content

### Articles With Images

```
1. Create directory reports/{article-name}/
2. Create images/ subdirectory
3. Download all images to images/
4. Generate Markdown (relative path references)

• Confidence: 80% confidence
• Finding: The skill persistently stores article content and downloaded images in workspace directories, which can leave sensitive browsing content resident beyond the immediate session. Persistent local artifacts increase exposure to later access by other tools, users, or processes on the same environment.

Static analysis

No suspicious patterns detected.