Back to skill

Security audit

story-short-write:短篇网文写作

Security checks for vulnerabilities and agentic risk

Overview

ClawScan classifies this writing skill as benign with medium confidence; the scanners found writing and author-memory features plus language and tool-scope notes, while analysis was partial.

Consider whether you are comfortable with the reported cross-session storage of author preferences and with detected file and shell capabilities. The scan was partial: 31 of 38 components were fully scanned and seven were only partially inspected, so the benign verdict is not a guarantee that every behavior was assessed.

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
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • 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
91% confidence
Finding

Author preferences can support fiction writing, but this chunk implements a durable author-memory subsystem rather than story ideation or drafting. The description does not disclose persistent profiling or cross-book memory management, which are substantive capabilities beyond ordinary writing implementation details. Notably, 'forget' retires an entry without erasing its stored assertion or evidence, and even queries may relocate misplaced state and regenerate files. No network communication or external exfiltration is evident.

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.

L131 requires the author-facing failure explanation to be in Chinese without user opt-in.

Medium
Category
Output Handling
Confidence
98% confidence
Finding

The instruction explicitly says to tell the author in Chinese what the design still lacks after unsuccessful repairs. It does not offer a language choice or explain why this operational message must be Chinese, even when the user prefers another language.

Content

No source excerpt is available for this finding.

L21 mandates Chinese for reporting conflicting delivery constraints without user opt-in.

Medium
Category
Output Handling
Confidence
98% confidence
Finding

The parameter-validation workflow explicitly instructs the agent to explain word-count and section-count conflicts in Chinese. This fixes the language of a user-facing operational message without offering a choice or documenting a reason that requires Chinese.

Content

No source excerpt is available for this finding.

The style guidance explicitly requires Chinese output without offering a user language choice.

Low
Category
Output Handling
Confidence
86% confidence
Finding

The instruction “主流宫斗 / 重生线仍走朴素中文” requires mainstream palace-intrigue and rebirth stories to use plain Chinese. Although it appears in Chinese-genre style guidance, it does not make the language requirement conditional on the user's requested language or explain why other output languages are excluded.

Content

No source excerpt is available for this finding.