Back to skill

Security audit

AI Drawing Prompt Engineering Expert

Security checks for vulnerabilities and agentic risk

Overview

This prompt-writing toolkit has no material security concerns, although two non-sexual examples contradict its adult-only promise.

Correct the two Doubao examples to use adult characters before reusing them, and specify your preferred prompt language. Reference images used with an external generation service will be shared with that service. The supplied adult-only declaration is not consistently reflected in the examples.

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
  • System Prompt LeakageDirect Leakage, Indirect Extraction, Tool-Based Exfiltration
  • 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 (4)

Model-family rules mandate English or Chinese prompt output without user opt-in.

Medium
Category
Output Handling
Confidence
97% confidence
Finding

The syntax table requires a “Full English paragraph” for FLUX models and a “Chinese natural-language paragraph” for domestic APIs. These are explicit output-language requirements without user choice or a documented technical or regional justification.

Content

No source excerpt is available for this finding.

The example generates prompts for a 16-year-old character despite the manifest's explicit adult-only restriction.

Medium
Category
Excessive Agency
Confidence
100% confidence
Finding

The example repeatedly supplies an explicit 16-year-old identity anchor, contradicting the skill's adult-only constraint. Because examples are intended for reuse and anchors are preserved verbatim, this inconsistency can propagate prohibited age values into generated prompts; however, the supplied illustration is non-sexual and does not itself demonstrate exploitation or sexualization.

Content

No source excerpt is available for this finding.

The example supplies prompts for a 16-year-old character despite the manifest's explicit adults-only restriction.

Medium
Category
Excessive Agency
Confidence
100% confidence
Finding

Version A explicitly defines a 16-year-old and carries that age through reusable prompt templates, violating the skill's adult-only constraint. Instructions to retain anchors verbatim can perpetuate this policy inconsistency, although the depicted campus portrait is non-sexual and the material does not demonstrate sexual exploitation.

Content

No source excerpt is available for this finding.

Indirect Prompt Extraction

Medium
Category
System Prompt Leakage
Confidence
75% confidence
Finding

Skill contains patterns that could indirectly extract system prompts through rephrasing, translation, summarization, or side-channel techniques.

Content

Scanner excerpt · references/ConversationalEditChain.csv (reported line 3)May include surrounding context.

text
ring garden, golden-hour backlight, 85mm half-body portrait.,P0 failure: anchor missing or rewritten → regenerate Round 1; do not proceed to Round 2 with broken anchors
Round 2,Single-point fix — feature,"Identify ONE feature that failed; describe the fix in one sentence; reference the anchor, do not rewrite it",Character identity anchors (mole / daisy / dress) frozen,Nanwang's dress should be mint green, not the current pale blue. Please keep the white daisy embroidery on the hem.,Fix failed → rephrase instruction using exact anchor wording: ""Change dress color to light mint #A8D8C9 as defined in anchor""
Round 3,Single-point fix — proportion,"Identify ONE proportion error; specify the correct ratio from AspectRatioBaselines; one sentence",Body proportions frozen at version baseline,The head-to-body ratio should be 6.5 heads (young woman baseline). Please adjust the proportions.,Fix failed → specify exact pixel/percentage: ""Reduce head size by 10% to match 6.5-head ratio""
Round 4,S