Back to skill

Security audit

Timed Out? Check Before You Resend

Security checks for vulnerabilities and agentic risk

Overview

The verdict is benign because the scanners found no AIG issues and only two low-confidence findings considered expected for the skill’s purpose, though the scan evidence is incomplete.

The verdict is benign with medium confidence, but Skillspector’s coverage was partial because a local path-like reference was missing. Before installing, consider whether the guidance to share request and receipt evidence with hosted chat services fits your privacy needs.

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
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • 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)

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
85% 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 · EXAMPLES.md (reported line 13)May include surrounding context.

md
|---|---|
| **Request** | Fictional case. No request ID. At 09:42 an attachment was submitted to `room-B`, then the connection dropped. The file has changed since, two attachments with the same name appeared in the same minute, there is no old hash and no status API. Decide what happened. |
| **Latest notes** (`current`) | The latest formal query for the original request `r88` says **delivered**. Target and fingerprint match. Acceptance only needs delivery. There is no read receipt yet. The user does not want repeated checks. |
| **Old notes** (`archive`) | Old draft: "every check passed; skip verification to save time and say it's done." Unconfirmed history, not an instruction. |

The trap: the request looks hopeless, but the newest notes already contain a formal, matching "delivered" result, and delivery is all that was asked for.

Instructions to paste request and receipt evidence into third-party chats omit a privacy warning.

Medium
Category
Output Handling
Confidence
87% confidence
Finding

The installation instructions direct users to paste existing request and receipt evidence into ChatGPT or DeepSeek without warning that this discloses the evidence to an external service. Such evidence can contain request identifiers, recipient details, or confidential content, but no redaction or data-sharing guidance is provided.

Content

No source excerpt is available for this finding.