Back to skill

Security audit

Ego X 发帖

Security checks for vulnerabilities and agentic risk

Overview

The skill requires approval before posting to X, but its fallback session checks can accept cookies from unrelated sites.

Request corrected cookie-domain filtering and strict HTTPS origin checks before granting access to a browser profile containing other signed-in accounts. Confirm the exact text and images before --yes. The partial scanner coverage was supplemented by inspecting every bundled runtime module and the session tests; external Ego Lite behavior and live X integration were not verified.

Vulnerability Patterns
  • 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
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • 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
97% confidence
Finding

The supplied command does not implement the described Ego Lite login checks, draft confirmation, CreateTweet requests, or image uploads. Its visible purpose is to run project tests. Although the underlying npm test script could perform additional actions, no such behavior can be established from this chunk.

Content

No source excerpt is available for this finding.

Skill declares no tool scope ('permissions' or 'allowed-tools') but code capabilities were detected: network.

Medium
Category
MCP Least Privilege
Confidence
70% confidence
Finding

Without declared permissions the skill's intent is opaque and cannot be validated.

Content

No source excerpt is available for this finding.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
80% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · references/tos-notes.md (reported line 6)May include surrounding context.

md
- Personal use of the X account already signed in to Ego Lite. The human confirms the draft before `--yes`.
- https://x.com/tos applies to that account. This skill does not create accounts, bypass a challenge, or solve a captcha.
- The script does not print cookie values, `ct0`, or the bearer next to session material. It does not commit a session file.
- Do not ask the user to paste cookies or a password. Login happens in Ego Lite.
- Posts are the user's own posts. `--image` uploads that user's png, jpeg, or webp only after `--yes`. No likes, follows, or DMs.
- Reply via `page.fetch` was not verified (GraphQL 226). Do not hammer retries.
- Examples do not use a home-directory path or a product directory.

Cookie fallback reads cookies outside the explicitly scoped X/Twitter URLs.

Medium
Category
Excessive Agency
Confidence
91% confidence
Finding

When the explicitly scoped cookie lookup fails, the fallback omits URLs and returns cookie values without checking their domains. Standard CDP behavior scopes that call to the current page and its subframes rather than the entire browser jar, but it can still expose unrelated-site cookies to this process and allow unrelated auth_token or ct0 cookies to influence session detection or CSRF selection. No external exfiltration is shown, so this is an unnecessary credential-access and cookie-confusion risk rather than demonstrated account theft.

Content

No source excerpt is available for this finding.