Back to skill

Security audit

AI短剧编剧(剧本分镜资产图全出)

Security checks for vulnerabilities and agentic risk

Overview

ClawScan rates this skill suspicious with medium confidence because scanners report automatic image generation and moderation-block recovery guidance, while part of the script analysis is incomplete.

Consider the reported automatic image-generation and download behavior, possible service costs, and the sharing of project prompts with an image service before installing. The scanner also flags moderation-block recovery guidance, and script analysis was partial; review these concerns and the incomplete coverage when deciding whether to install.

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
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (7)

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 105)May include surrounding context.

md
进表),裸切零;爽文向打脸三件套(嚣张镜→打脸镜→围观反应镜)逐次打脸逐套核;收尾全表过一遍过测试(见「镜头因果链纪律」节)。**出表后先机器核数:`node scripts/storyboard-check.js <项目目录>/分镜表.md`——有FAIL照单回炉(缺承接句补/台词超20字砍/切点行数补齐/超8秒拆镜

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 170)May include surrounding context.

md
系≥3套|引擎三选一|声部≥5|点≥3|冲突对≥3|爽阀≥3中段|互动回合≥8物理≥4|逻辑链裸切零|媒体≥2|母题首尾回声|揭示≥2阶段。数字可机检的部分由 scripts/storyboard-check.js 把关。

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 193)May include surrounding context.

md
□ 丰富度:**贴 `node scripts/storyboard-check.js` 完整输出(须 exit 0)**;预算十五条对照 richness-standard.md 逐项达标(不达标项列出豁免理由)

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 213)May include surrounding context.

md
| scripts/storyboard-check.js | 分镜表机器核数:预算vs实测逐项输出,FAIL回炉(零依赖,node直接跑) |

The workflow defaults to generating and downloading images, while the manifest advertises image prompts.

Medium
Category
Excessive Agency
Confidence
92% confidence
Finding

The advertised deliverables include image prompts, but this workflow automatically submits image-generation jobs and downloads results unless the user explicitly opts out. That expands a writing task into service use and filesystem changes without a clear authorization checkpoint, potentially consuming paid resources or creating unwanted artifacts; the creative-production context makes generation relevant but does not establish consent.

Content

No source excerpt is available for this finding.

Automatic image generation sends project prompts to an image-generation service without a privacy warning.

Medium
Category
Output Handling
Confidence
88% confidence
Finding

Automatic submission sends project-derived prompts to an image-generation service without a stated privacy review or disclosure checkpoint. Unreleased scripts, proprietary character descriptions, or identifying details could therefore cross a service boundary unexpectedly, although the supplied source does not establish the provider's retention practices or show deliberate exfiltration.

Content

No source excerpt is available for this finding.

Moderation-block recovery instructs users to resume blocked material in a new conversation using euphemistic framing.

Medium
Category
Prompt Injection
Confidence
93% confidence
Finding

The FAQ recommends starting a new conversation, restoring the project from progress.md, and using euphemistic framing after a moderation block. This can preserve the blocked task while removing conversational context and obscuring its substance from safeguards; nearby content restrictions and a feedback option suggest troubleshooting intent but do not eliminate that bypass risk.

Content

No source excerpt is available for this finding.