T09 · Insecure Skill Coding Practices
Warning
- Location
roles/02-product.md:17- Finding
Rewriting Rules Can Introduce Unsupported First-Hand Claims and Fabricated Details
- Content
View full analysis
- Remediation
View remediation
Security audit
超能去AI
Security checks for vulnerabilities and agentic risk
Overview
This is not malware, but it is designed to make AI text appear human-written and can add unsupported personal or factual details, so it belongs in Review before installation.
Install only if you specifically want Chinese creative or informal text to be heavily humanized. Avoid using it for academic, legal, news, technical, product, workplace, scientific, compliance, or authorship-sensitive writing unless every generated detail is checked and removed when unsupported. Consider narrowing activation triggers to avoid accidental rewrites.
Vulnerability Patterns
- Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
- 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
Findings (1)
Vulnerability Patterns
- Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
- Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
- 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 (17)
Description-Behavior Mismatch
High
- Category
- Not specified by scanner
- Confidence
- 98% confidence
- Finding
The skill claims it will preserve facts and core meaning, but its required workflow and hard constraints instruct the model to add private-feeling details, distractions, and narrative elements not present in the source. That creates a built-in risk of fabrication and misrepresentation, especially when users may rely on the output as a faithful rewrite rather than partially invented content.
- Content
Ssd 4
High
- Category
- Not specified by scanner
- Confidence
- 99% confidence
- Finding
The hard constraints explicitly require injecting errors, dead ends, private-feeling sensory associations, and other artifacts intended to simulate authentic human writing. This is a deliberate deception pattern: it does not merely restyle text, it manufactures signals of human origin that can mislead readers, reviewers, or detectors about provenance and authenticity.
- Content
Vague Triggers
Medium
- Category
- Not specified by scanner
- Confidence
- 94% confidence
- Finding
The trigger list contains common conversational phrases such as '帮我改一下，太像AI写的了' and '这看着太AI了' that could plausibly appear in ordinary user text, increasing the chance of unintended skill activation. In a rewriting skill, accidental invocation can cause unrequested transformation of user content and may override the user's actual task flow, especially when the skill is designed to materially alter wording and even add details.
- Content
Natural-Language Policy Violations
Medium
- Category
- Not specified by scanner
- Confidence
- 88% confidence
- Finding
SQP-3 适用于所有文件类型。描述中写明“专为中文文本设计”，属于语言/locale 限制，但未见用户可选择其他语言的机制，也未说明这是出于地域合规或其他必要的组织性约束，构成潜在的语言策略违例。
- Content
Vague Triggers
Medium
- Category
- Not specified by scanner
- Confidence
- 91% confidence
- Finding
The trigger phrases are broad, colloquial expressions that overlap with normal user requests, so the skill may activate unintentionally. In this skill's context, accidental activation is more dangerous because the behavior is not simple paraphrasing—it systematically alters style, structure, and potentially content to disguise AI authorship.
- Content
Ssd 4
Medium
- Category
- Not specified by scanner
- Confidence
- 95% confidence
- Finding
The skill's stated purpose is to transform text so it appears human-authored across language, narrative, detail, and persona. That is dangerous because it is explicitly optimized to conceal AI generation, which can facilitate deception, provenance evasion, academic or professional misrepresentation, and circumvention of disclosure norms.
- Content
Ssd 4
Medium
- Category
- Not specified by scanner
- Confidence
- 94% confidence
- Finding
The workflow normalizes a staged process for adding humanizing texture and narrative reshaping, including role loading and shared techniques, which operationalizes deception rather than incidental style improvement. The skill context makes this more dangerous because the cumulative steps are designed to systematically disguise machine authorship at scale.
- Content
Intent-Code Divergence
Medium
- Category
- Not specified by scanner
- Confidence
- 91% confidence
- Finding
The file says factual documents should not be rewritten and should be left as-is, but elsewhere defines rewrite roles and processes for technical, business, workplace, and science content. This contradiction increases the chance that factual material will still be transformed by an agent, causing accidental distortion of data, claims, or source structure.
- Content
Natural-Language Policy Violations
Medium
- Category
- Not specified by scanner
- Confidence
- 93% confidence
- Finding
The instruction says that when the source text's tone is unclear, the editor should default to '平实中文', which imposes a specific language choice. This is a natural-language policy concern because it does not offer the user a language or locale opt-in, nor does the file justify a Chinese-only scope.
- Content
Natural-Language Policy Violations
Medium
- Category
- Not specified by scanner
- Confidence
- 92% confidence
- Finding
The entire role specification is written as a prescriptive persona in Chinese, with no indication that the user may choose another language or that the skill is intended only for a Chinese-language context. Under the policy, a skill that effectively enforces a specific language without opt-in is a natural-language policy violation.
- Content
Natural-Language Policy Violations
Medium
- Category
- Not specified by scanner
- Confidence
- 95% confidence
- Finding
The entire role definition prescribes Chinese output and Chinese-specific wording rules, including mandatory replacements and examples, without stating that the user can choose another language. This can violate language/locale policy when a skill imposes a specific language by default rather than offering opt-in or choice.
- Content
Natural-Language Policy Violations
Medium
- Category
- Not specified by scanner
- Confidence
- 94% confidence
- Finding
The entire role specification is written as a Chinese-language persona and style guide, with no indication that users may opt into another language or locale. Under the policy, forcing a specific language without user choice is a natural-language policy violation unless the locale constraint is explicitly documented and justified.
- Content
Natural-Language Policy Violations
Medium
- Category
- Not specified by scanner
- Confidence
- 94% confidence
- Finding
This markdown file is entirely framed as a Chinese copywriting role and instructs stylistic behavior only in Chinese, with no indication that users may opt into another language. Under the policy, forcing a specific language without user opt-in is a natural-language policy violation.
- Content
Natural-Language Policy Violations
Medium
- Category
- Not specified by scanner
- Confidence
- 94% confidence
- Finding
The entire role definition is written as prescriptive guidance in Chinese and presents a fixed Chinese social-media voice without offering any language or locale choice. Under the stated policy, forcing a specific language without user opt-in is a natural-language policy violation unless the locale constraint is explicitly documented and justified.
- Content
Natural-Language Policy Violations
Medium
- Category
- Not specified by scanner
- Confidence
- 95% confidence
- Finding
The file is entirely written as a role specification in Chinese and defines tone, phrasing, and rewrite behavior accordingly, but it does not state that language choice is optional or limited to a China-specific context. Under the policy, enforcing a specific language without user opt-in is a natural-language policy violation.
- Content
Natural-Language Policy Violations
Medium
- Category
- Not specified by scanner
- Confidence
- 85% confidence
- Finding
The file title and opening instructions are entirely framed as universal guidance in Chinese for all roles, and there is no indication that users may choose another language or that the locale restriction is necessary for a region-specific purpose. This can violate language/locale policy because it imposes a single language by default rather than offering opt-in or documenting justification.
- Content
Natural-Language Policy Violations
Low
- Category
- Not specified by scanner
- Confidence
- 90% confidence
- Finding
The README states "专为中文文本设计", which enforces a locale/language limitation in the skill description. The file does not present this as a user choice or provide a region/compliance justification for the restriction.
- Content