Back to skill

Security audit

LinkedIn: post & schedule posts (1Social)

Security checks for vulnerabilities and agentic risk

Overview

The skill clearly describes social publishing through 1Social and requires approval before posting, deleting published posts, or uploading local files.

Connect only intended accounts and verify every destination before approving a post. Upload only files suitable for public sharing; cancelling a post does not remove uploaded media. Some networks require manual deletion. Review [1Social's privacy policy](https://1social.dev/legal/privacy) for its handling of tokens, posts, and media.

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 Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (2)

The LinkedIn-only manifest is expanded into multi-network publishing.

Medium
Category
Excessive Agency
Confidence
98% confidence
Finding

The manifest at L03 describes publishing and managing LinkedIn posts, but L63–L65 explicitly instruct the agent to publish simultaneously to Instagram, TikTok, X, Threads, Facebook Pages, YouTube Shorts and Bluesky by selecting multiple channel IDs. Publishing to additional networks changes the destinations and audience of public writes; it is not an implementation detail of LinkedIn posting.

Content

No source excerpt is available for this finding.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
75% 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 · SKILL.md (reported line 74)May include surrounding context.

md
- **Unschedule:** `cancel_post`. Use this, not `delete_post`, for a post that has not been published.
- **Delete a published post:** `delete_post`, only after the user confirms. Report per network what actually came down: Instagram, TikTok, Threads and YouTube posts stay live and only the user can remove them.
- **A delivery failed:** `retry_delivery` retries one network and spends quota again. It is refused for permanent failures such as a revoked connection — then the user must reconnect at https://1social.dev/app.
- **A delivery is `unknown`:** ask the user to look at the network. Only when they tell you whether the post is live, call `confirm_delivery` with their answer. Never guess, and never `retry_delivery` with `force` without asking — it may publish a second copy.