Back to skill

Security audit

Ai Connect

Security checks for vulnerabilities and agentic risk

Overview

This skill is a web-reading helper, but it asks users to export login cookies without explaining secure handling or the account-takeover risk.

Review this skill before installing if you plan to use authenticated platforms. Do not paste exported cookies into chat or logs; treat them like passwords, use platform OAuth or scoped tokens where available, restrict any local cookie file permissions, and revoke sessions if cookies may have been exposed.

Vulnerability Patterns
  • 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
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
Findings (1)

T09 · Insecure Skill Coding Practices

Warning
Location
SKILL.md:111
Finding

Unsafe Handling of Exported Browser Session Cookies

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, lines 111-127
Vulnerability Type: Exposure and insecure handling of authentication cookies
Risk Level: Medium

Vulnerable Documentation Segment

The following is an English rendering of the complete affected segment:

text
## Platforms Requiring Configuration

Some platforms require login cookies (tell the Agent, "Help me configure XXX"):

| Platform | Configuration method |
|----------|----------------------|
| Twitter | Cookie |
| Xiaohongshu | Cookie |
| Reddit | rdt login |
| LinkedIn | Cookie |

Configuration process:
1. Log in to the target website through a browser.
2. Export the cookies with the Cookie-Editor extension.
3. Tell the Agent, "Help me configure Twitter, Xiaohongshu, etc."

Technical Analysis

Exported browser cookies can contain active authentication session tokens. Possession of such a token may allow a party to use the corresponding account without knowing the password or completing the normal authentication flow.

The documented workflow directs users to export cookies through a browser extension and then involve the Agent in configuring the affected service. It does not specify a secure secret-input channel, approved storage destination, restrictive file permissions, log redaction, expiration policy, cleanup process, or session-revocation procedure.

Consequently, a user may paste cookie contents into an Agent conversation or allow them to appear in tool arguments, command history, execution traces, logs, or insecure configuration files. This creates an avoidable credential-disclosure risk.

Attack Path

  1. A user signs in to Twitter, Xiaohongshu, LinkedIn, or another supported service.
  2. The user exports the authenticated browser cookies with the recommended browser extension.
  3. Following the Skill instructions, the user provides those cookies to the Agent or asks the Agent to store them during confi ...[truncated 1211 chars]
Remediation
View remediation

Remediation Suggestions

  1. Do not instruct users to paste cookie values into Agent conversations, prompts, or ordinary command arguments.
  2. Use a documented local secret store or a dedicated masked secret-input mechanism that prevents credentials from entering chat history and logs.
  3. Clearly identify the trusted tool and exact local destination receiving the cookies before requesting any configuration.
  4. Store cookie files with permissions limited to the owning user, such as mode 0600 on supported systems.
  5. Disable command echoing and redact cookie values from tool output, diagnostic reports, error messages, telemetry, and execution traces.
  6. Request only the minimum platform-specific cookies necessary for the feature instead of complete browser cookie exports.
  7. Never reuse a general browser profile or export unrelated cookies from other domains.
  8. Define retention and deletion behavior, including secure removal when configuration is replaced or no longer required.
  9. Warn users that cookies are authentication credentials and must not be shared through chat, tickets, screenshots, or public files.
  10. Provide instructions for revoking active sessions and rotating credentials after actual or suspected exposure.
  11. Prefer platform-supported OAuth or restricted API tokens with explicit scopes, short lifetimes, and revocation support wherever available.
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (2)

Missing User Warnings

High
Category
Not specified by scanner
Confidence
97% confidence
Finding

The skill instructs users to export browser cookies and provide them to the agent for multiple platforms, but it does not warn that cookies are highly sensitive authentication material equivalent to active session credentials. This can lead users to disclose reusable login tokens to the agent or surrounding infrastructure, enabling account takeover, impersonation, or unauthorized access if those cookies are exposed, logged, or mishandled.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The trigger phrases are very broad everyday terms such as '联网', '看视频', and '读帖子', which can cause the skill to activate in contexts where the user did not explicitly intend to invoke this capability. Because the skill performs network retrieval and external tool usage, accidental invocation expands the agent's attack surface and can lead to unintended data access, external requests, or unsafe tool chaining.

Content

No source excerpt is available for this finding.