Back to skill

Security audit

story-import:逆向导入已有小说

Security checks for vulnerabilities and agentic risk

Overview

ClawScan labels this skill benign with medium confidence: its findings mainly describe a mismatch between the advertised novel-import workflow and the code that was inspected, alongside some scope and validation concerns; this is not a claim that it is risk-free.

The verdict is benign, but confidence is medium and the scan is partial: 18 of 19 components were scanned, one file was only partly inspected, and a referenced local path was missing. Before installing, consider whether the reported outline-viewing and validation behavior matches your needs, and whether you are comfortable with the undeclared file and shell capabilities and broad activation trigger.

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
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • 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 (5)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
88% confidence
Finding

The supplied chunk implements an outline-selection utility, not the declared novel-import workflow. Its inputs are already-structured outline files, and its outputs are terminal text and diagnostics. It could support a larger import pipeline, but that connection is not demonstrated here. This assessment is limited to the supplied chunk; it shows no network access, file modification, or other suspicious side effects.

Content

No source excerpt is available for this finding.

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

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.

The activation list includes the unqualified trigger “导入” (“import”), which is broader than novel-project importing.

Medium
Category
Trigger Abuse
Confidence
92% confidence
Finding

“导入” is commonly used for importing data, documents, settings, and other unrelated content. Although the description explains the novel-import purpose, it explicitly lists this generic word as a trigger without requiring a novel-related context or providing exclusions.

Content

No source excerpt is available for this finding.

The verifier enforces value-level acceptance rules despite documenting structural-only validation that permits unknown values.

Low
Category
Excessive Agency
Confidence
96% confidence
Finding

L010-L013 explicitly state that the verifier never judges values and that unknown values may be represented by [待补充]. However, L371-L386 make a recognized numeric word-count target within enforced bounds a blocking requirement; L243-L258 also explicitly reject placeholders for two required fields. Consequently, an imported outline following the documented unknown-value policy can fail verification.

Content

No source excerpt is available for this finding.

Undeclared outline sections are silently promoted to volume-wide scope, contradicting the documented warning behavior.

Low
Category
Excessive Agency
Confidence
99% confidence
Finding

The module documentation at L004 and L012 says undeclared sections remain undeclared and are conservatively included with a warning, including unannotated children of volume-wide sections. Instead, when any section uses inline scope syntax, this block assigns volume-wide scope to every section lacking both a scope and unit identifier. Those sections consequently bypass the undeclared-section warnings and strict validation at L306–L309, and are treated as permanent material in subsequent selections.

Content

No source excerpt is available for this finding.