照片分析与 Darktable 修图

• Outcome: Pass
• Latest audit: Sep 16, 2026 · 10:20 AM UTC
• Version: 1.0.0

Overview

This Markdown-only Darktable photo-editing skill is coherent and purpose-aligned, with only minor disclosure and language-formatting caveats.

Installers should understand that the skill may ask the agent to inspect provided photo files and metadata, and may use official Darktable or museum web sources when current-version or publication-ready facts matter. In air-gapped use, treat the live-verification parts as unavailable and expect the agent to state the evidence gap.

A.I.G

Vulnerability Patterns

• Skill Instruction Hijacking: Alters the agent's session goals or safety constraints when the skill loads
• Agent Memory Poisoning: Writes attacker-controlled rules into memory that affect later sessions
• Remote Payload Retrieval and Execution: Fetches external code whose behavior can change after review
• Embedded Malicious Code: Ships malicious scripts inside the skill and executes them locally
• Unauthorized Access and Privilege Escalation: Obtains permissions beyond the task's legitimate needs

SkillSpector

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Memory Poisoning: Persistent Context Injection, Context Window Stuffing, Memory Manipulation
• Rogue Agent: Self-Modification, Session Persistence
• MCP Tool Poisoning: Hidden Instructions, Unicode Deception, Parameter Description Injection
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands

Findings (7)

Self-Modification

High

• Category: Rogue Agent
• Confidence: 85% confidence

Finding

Skill modifies its own code, configuration, or behavior at runtime. Self-modification enables an agent to escalate privileges, disable safety constraints, or install persistent backdoors.

Content

Scanner excerpt · references/version-source-audit.md (reported line 68)May include surrounding context.

md

2. Open the current official HTML page for the module or preference.
3. Check the installed version's official release notes for additions/removals.
4. Compare the user's screenshot with the current page.
5. Update this skill only after confirming a durable change; do not encode a one-off packaging bug as a general rule.

Description-Behavior Mismatch

Medium

• Category: Not specified by scanner
• Confidence: 94% confidence

Finding

The manifest states the skill has 'Zero dependencies' and 'works offline', but the body later requires live web verification in some scenarios. This can mislead users, policy engines, or deployment controls into assuming the skill never needs network access, creating trust and operational mismatches around privacy, reproducibility, and sandboxing.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium

• Category: Not specified by scanner
• Confidence: 94% confidence

Finding

The documentation advertises offline operation, yet the skill instructs the agent to use web tools and verify against official manuals and release notes when discrepancies arise. In security-sensitive or air-gapped contexts, this discrepancy can cause unintended network use or bypass assumptions made by users and platform administrators about data locality and external communication.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium

• Category: Not specified by scanner
• Confidence: 96% confidence

Finding

The file embeds a compact output template that explicitly requires Chinese-language headings and a fixed fallback sentence, which can override or conflict with the user's requested language. In an agent skill, this is a prompt-quality and policy issue because it can cause unauthorized language switching and reduce usability or trust, though it does not create code execution or data exfiltration risk.

Content

No source excerpt is available for this finding.

Context Window Stuffing

Medium

• Category: Memory Poisoning
• Confidence: 85% confidence

Finding

Skill attempts to fill the context window with filler content, displacing legitimate instructions and safety constraints. This can degrade agent performance or bypass safety boundaries.

Content

Scanner excerpt · references/interface-navigation.md (reported line 126)May include surrounding context.

md

Useful default shortcuts:

- Tab: temporarily hide panels and fill the window with the center view.
- F11: full screen.
- Shift+Ctrl+T: top panel.
- Shift+Ctrl+B: bottom panel.

Natural-Language Policy Violations

Low

• Category: Not specified by scanner
• Confidence: 87% confidence

Finding

The instructions require module labels in both Chinese and English, including exact English UI names. This imposes a specific language/locale format on responses without indicating user choice or opt-in, which matches the language/locale policy category.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low

• Category: Not specified by scanner
• Confidence: 92% confidence

Finding

Line L13 instructs that the English module name must be used, which is a language-specific requirement stated as mandatory. The file does not provide an opt-in, alternative locale option, or justification that the skill is restricted to an English-only interface context.

Content

No source excerpt is available for this finding.