Back to skill

Security audit

Ego WeChat Channels Video

Security checks for vulnerabilities and agentic risk

Overview

Review before installing: the upload helper can expose unpublished videos to unrelated web pages, and the optional originality step accepts an agreement without a separate approval check.

Prefer a version that restricts the file helper to the intended origin, requires an unpredictable access token, and shuts down automatically. Until then, avoid confidential videos and stop the helper immediately after transfer. Leave originality disabled unless you have reviewed and authorized its agreement. Inspection covered every bundled file, but did not test live Ego or WeChat behavior.

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
scripts/serve-file.mjs
Finding

Unauthenticated wildcard-CORS server exposes the selected video to unrelated origins

Content
View full analysis

Impact

An attacker-controlled page can read and transmit the unpublished video while the helper is running. Exposure is limited to the selected file; this is not arbitrary filesystem access.

Vulnerability Patterns
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • YARA SignaturesMalware Match, Webshell Match, Cryptominer Match
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (4)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding

The supplied chunk contains only npm test, with no visible login check, browser interaction, video upload, metadata entry, or draft saving. Its observable purpose is test execution, materially different from the declared workflow. Any additional behavior of the underlying npm script cannot be determined from this chunk.

Content

No source excerpt is available for this finding.

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 66)May include surrounding context.

md
node scripts/serve-file.mjs --file /absolute/path/to/video.mp4 --port 18765

YARA rule 'info_stealer': Information stealer patterns (credential harvesting, browser data theft) [malware]

High
Category
YARA Match
Confidence
75% confidence
Finding

YARA rule matched a known malware signature (reverse shell, backdoor, ransomware, C2 framework, or info stealer).

Content

Scanner excerpt · scripts/test/pure.test.mjs (reported line 71)May include surrounding context.

js
rt, 9);
  assert.equal(parseServeArgs(["--help"]).help, true);
  assert.equal(parseServeArgs(["-h"]).file, null);
  assert.throws(() => parseServeArgs(["--file", "/tmp/a.mp4", "--port", "0"]), /--port/);
  assert.throws(() => parseServeArgs(["--file", "/tmp/a.mp4", "--port", "nope"]), /--port/);
  assert.throws(() => parseServeArgs(["--port"]), /--file is required/);
});

test("classifyLogin uses cookies, draft chrome, and ignores QR text off the login page", () => {
  const cookie = classifyLogin({
    url: "https://channels.weixin.qq.com/",
    signals: { loggedInCookie: true },
  });
  assert.equal(cookie.logged_in, true);
  assert.equal(cookie.has_qr_hint, undefined);
  const draft = classifyLogin({ url: "https://channels.weixin.qq.com/platform/post/list", text: "草稿箱" });
  assert.equal(draft.logged_in, true);
  const query = classifyLogin({ url: "https://channels.weixin.qq.com/login?from=home", text: "微信扫一扫" });
  assert.equal(query.on_login_page, true);
  assert.e

The originality workflow accepts a platform agreement on the user's behalf without warning or obtaining informed authorization.

Medium
Category
Output Handling
Confidence
92% confidence
Finding

The optional originality workflow directs the agent to check an agreement and confirm an originality declaration without explicitly requiring informed user authorization. Saving only a draft reduces public-exposure risk, but does not establish permission to accept platform terms or make rights-related representations on the user's behalf.

Content

No source excerpt is available for this finding.