Security audit
Security checks for vulnerabilities and agentic risk
This is a coherent Chinese short-fiction writing skill with local drafting, review, and preference-memory helpers, with no evidence of deception, exfiltration, remote code, or destructive behavior.
Install this only in writing workspaces where you want the agent to create and modify story files, run local review scripts, optionally use configured writing subagents, and keep project-local author preference memory. Review sensitive genre outputs before sharing, especially revenge, privacy-abuse, or violent fiction themes.
Reading local configuration such as .deslop-whitelist and applying whitelist-based masking is outside the declared trigger and purpose of a fiction-writing skill. Even if intended for quality control, undisclosed config access expands the local attack surface and can reveal project conventions or hidden files to the agent workflow.
Reading local configuration such as .deslop-whitelist and applying whitelist-based masking is outside the declared trigger and purpose of a fiction-writing skill. Even if intended for quality control, undisclosed config access expands the local attack surface and can reveal project conventions or hidden files to the agent workflow.
Reading local configuration such as .deslop-whitelist and applying whitelist-based masking is outside the declared trigger and purpose of a fiction-writing skill. Even if intended for quality control, undisclosed config access expands the local attack surface and can reveal project conventions or hidden files to the agent workflow.
Reading local configuration such as .deslop-whitelist and applying whitelist-based masking is outside the declared trigger and purpose of a fiction-writing skill. Even if intended for quality control, undisclosed config access expands the local attack surface and can reveal project conventions or hidden files to the agent workflow.
Reading local configuration such as .deslop-whitelist and applying whitelist-based masking is outside the declared trigger and purpose of a fiction-writing skill. Even if intended for quality control, undisclosed config access expands the local attack surface and can reveal project conventions or hidden files to the agent workflow.
Reading local configuration such as .deslop-whitelist and applying whitelist-based masking is outside the declared trigger and purpose of a fiction-writing skill. Even if intended for quality control, undisclosed config access expands the local attack surface and can reveal project conventions or hidden files to the agent workflow.
Reading local configuration such as .deslop-whitelist and applying whitelist-based masking is outside the declared trigger and purpose of a fiction-writing skill. Even if intended for quality control, undisclosed config access expands the local attack surface and can reveal project conventions or hidden files to the agent workflow.
Reading local configuration such as .deslop-whitelist and applying whitelist-based masking is outside the declared trigger and purpose of a fiction-writing skill. Even if intended for quality control, undisclosed config access expands the local attack surface and can reveal project conventions or hidden files to the agent workflow.
Referenced artifact was not completely inspected
- [ ] `node scripts/check-ai-patterns.js --check --fail-on=blocking 正文.md` 无 blocking 命中；其余提示先通读，确属问题再改
Referenced artifact was not completely inspected
- [ ] `node scripts/check-degeneration.js --check 正文.md` 无 blocking 退化命中（复读/截断/工程词泄漏）
The skill explicitly instructs the agent to read, write, and invoke shell commands across project files, but it declares no tool scope or permission boundaries. That creates an avoidable least-privilege failure: a writing skill can modify local content and run local checks without any user-visible constraint, increasing the chance of unintended file access or destructive edits.
The trigger set includes broad everyday phrases such as '帮我写一篇短篇' and '写个盐言故事', which can cause accidental activation outside a deliberate skill invocation. In a skill with file and shell capabilities, unintended activation is more dangerous because routine conversation may launch multi-step workspace reads, writes, or checks.
The description, prompts, and workflow are entirely written to operate in Chinese, including prescribed user-facing prompts and output artifacts, but the file does not state that Chinese is optional or limited to a justified region-specific use case. This can violate language/locale policy when a skill implicitly forces one language without user opt-in.
Lines L066-L069 prescribe exact reply wording entirely in Chinese and require the assistant to respond in that format, but the document does not provide any user language choice or justify a locale restriction. This is a natural-language policy issue because it forces a specific language for user-visible interaction.
This markdown file presents all instructions and policy guidance exclusively in Chinese, and there is no indication that users may choose another language or that the skill is intentionally limited to a Chinese-language audience. Under the language/locale policy, forcing a specific language without user opt-in is a natural-language policy concern.
This markdown file defines activation behavior, and the phrase '类别不明时检查上表全部表达与用笔项' broadens execution when the caller's category is unclear. That fallback lacks clear constraints or negative examples, which could cause unintended invocation of broader checks than the user intended.
SQP-3 applies to all file types and covers natural-language policy violations such as forcing a specific language without user opt-in. The title and all instructions are written only in Chinese, with no indication that users may choose another language or locale.
The entire skill is written in Chinese and directly instructs the user in that language, but it does not state that Chinese is optional or that the skill is intentionally limited to a Chinese-language context. Under the policy, forcing a specific language without user opt-in is a natural-language policy violation.
This file’s natural-language instructions and examples are exclusively Chinese, and there is no indication that the skill is limited to Chinese-speaking users or that users can opt into another language. Under the language/locale policy, forcing a specific language without user choice can be a policy violation.
All instructions, examples, and style guidance are written as mandatory Chinese-language output conventions, with no indication that the user may choose another language. This creates a language/locale policy issue unless the file explicitly documents that the skill is intentionally limited to Chinese-only use for a justified regional context.
This style guide explicitly promotes plots centered on non-consensual intimate-image sharing, doxxing, and public humiliation as high-impact dramatic devices, but provides no safety framing or constraints for handling sexual privacy abuse. In a writing skill, that can normalize or encourage generation of revenge-fiction content built around image-based abuse and privacy violations, increasing the chance the model produces harmful or exploitative material without guardrails.
The document endorses extreme violence themes such as dismemberment, crushing, and other severe retaliatory harm as genre selling points, while specifically advising authors to preserve emotional intensity. Even though framed as fiction guidance, the lack of safety boundaries makes the skill more likely to generate graphic revenge narratives or escalate user prompts toward violent catharsis.
The instruction states that mainstream palace-intrigue/rebirth lines must use 'plain Chinese' and avoid modern internet slang, which imposes a specific language/locale style as a requirement. This is a natural-language policy concern because the file does not present this as an optional preference or offer any user choice.
The file title and all instructional content are written exclusively in Chinese, with no indication that the skill is region-specific or that users may opt into another language. Under the language/locale policy, forcing a specific language without user opt-in is a natural-language policy violation.
This markdown file presents all instructions in Chinese and does not indicate that the user can choose another language or that the skill is intentionally limited to Chinese-language users. Under the stated policy, forcing a specific language without user opt-in is a natural-language policy violation.
No suspicious patterns detected.