Back to skill

Security audit

Sci Translation Polish

Security checks for vulnerabilities and agentic risk

Overview

This is a text-only academic translation and polishing skill with some routing and install cautions, but no hidden execution, persistence, credential use, or destructive behavior found.

Install only from the verified publisher/source, avoid running unpinned npx commands in sensitive environments, and use the skill only on manuscript content you are authorized to share with your agent. Confirm the target journal, target language, and scope before letting it rewrite full papers or reviewer correspondence.

Vulnerability Patterns
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • 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)

T08 · Insecure Dependencies

Warning
Location
README.md:21
Finding

Unpinned npx Installation Executes Mutable Third-Party Dependencies

Content
View full analysis
/sci-translation-polish ``` ``` `README_CN.md:93-98`: ```markdown ### Installation ```bash npx skills add xioahu1232/sci-translation-polish ``` ``` ### Technical Analysis The documented installation procedure invokes `npx`, which can download and execute a package resolved from an external package registry. The `skills` CLI is not pinned to a reviewed version, and the referenced skill is not tied to an immutable release or commit hash. Consequently, the code executed by the command can differ from the code that was reviewed during this audit. This creates a supply-chain trust boundary in which package-owner compromise, registry compromise, package takeover, or a malicious future release could introduce attacker-controlled installation behavior. The English example additionally uses an unspecified `` placeholder. If users replace it with an incorrect, untrusted, or typosquatted package owner, they may install a different skill than intended. The repository itself is text-only and does not automatically execute this command. Exploitation requires a user to follow the installation documentation while the remotely resolved dependency or skill source is malicious. ### Attack Path 1. An attacker compromises the package publisher, remote skill repository, registry account, or another source resolved by the installation command. Alternatively, a user selects a malicious or typosquatted owner when replacing the `` placeholder. 2. The attacker publishes a modified version containing malicious installation or runtime behavior. 3. A user runs the documented unpinned `npx skills add ...` command. 4. `npx` retrieves the currently resolved `skills` package, and the CLI r ...[truncated 1081 chars]
Remediation
View remediation
add ... ``` 2. Pin the skill itself to an immutable version, release digest, or commit hash rather than resolving the latest remote state. 3. Replace the `` placeholder with the verified canonical publisher name in all documentation. 4. Document the expected registry and canonical repository so users can detect typosquatted or substituted sources. 5. Publish and verify cryptographic integrity information, such as release checksums, signed tags, provenance attestations, or package signatures. 6. Configure installation tooling to reject unsigned, unexpected, or mutable sources where supported. 7. Advise users not to run the installation command with `sudo` or an administrator account. 8. Recommend reviewing the resolved package version and skill contents before installation, particularly in CI systems or environments containing sensitive credentials. 9. Keep the English and Chinese installation instructions synchronized so both use the same pinned and verified source. ]]>
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Rogue AgentSelf-Modification, Session Persistence
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (29)

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The workflow explicitly instructs the model to produce English output without checking whether the user requested English or wants bilingual preservation. This can override user intent and cause unwanted disclosure or transformation of user-provided academic content, especially when the user only asked for analysis, polishing guidance, or partial assistance rather than full translation.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
70% confidence
Finding

npx commands without a version suffix (e.g. @1.0.0) create a rug-pull risk if the upstream server is compromised and publishes a malicious update.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
70% confidence
Finding

npx commands without a version suffix (e.g. @1.0.0) create a rug-pull risk if the upstream server is compromised and publishes a malicious update.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The usage examples and metadata encourage invocation on very broad phrases like 'Make this publication-ready' and instruct use whenever users mention paper writing or journal submission, even without explicitly requesting translation. That can cause unintended auto-activation, pulling sensitive manuscript content into this skill when the user may have wanted general advice, which increases prompt-routing and data-exposure risk.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The README instructs users to 'think in English' and 'write in English' and frames the skill as converting Chinese papers into publication-grade English only. Because SQP-3 applies to all file types, this fixed language constraint is a natural-language policy concern unless the skill offers a user choice or clearly documents a justified locale restriction.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

This line reinforces a mandatory English-only behavior: '不是翻译,是跨语言学术写作。用英语思考,用英语写作。' Under SQP-3, forcing a specific language without user choice or clear justification can be a policy violation.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The skill explicitly instructs automatic use whenever a user mentions broad academic-writing scenarios, even if they do not ask for translation or polishing. This can cause inappropriate routing of user requests, override user intent, and expose sensitive manuscript content to unnecessary transformation steps or disclosures within the skill workflow.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The skill description mandates English output for international publication without offering user language preference or fallback behavior. While aligned with the skill's purpose, this can still mis-handle multilingual or bilingual requests and produce undesired output that conflicts with user expectations or workflow requirements.

Content

No source excerpt is available for this finding.

Session Persistence

Medium
Category
Rogue Agent
Confidence
75% confidence
Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · examples/education_examples.md (reported line 65)May include surrounding context.

md
With the in-depth advancement of the popularization process of higher education, the issue of teaching quality assurance in colleges and universities has increasingly become the focus of academic and policy makers' attention. This paper takes a provincial university as a case, adopting fuzzy comprehensive evaluation method to conduct a systematic evaluation of its undergraduate teaching quality. The research found that the overall teaching quality of this university is at a "good" level, but there are significant differences in different dimensions. Specifically, faculty team construction and teaching condition guarantee scored higher, while classroom teaching quality and student development support scored relatively lower. Further analysis showed that key factors affecting teaching quality include: imperfect teacher teaching ability development mechanism, curriculum setting disconnected from market demand, teaching evaluation system too single, etc. Based on the above findings, this paper suggests that colleges and universities should establish a normalized teaching quality monitoring mechanism, improve the teacher teaching ability development support system, and actively promote student-centered teaching model reform.

### ✅ Native English
As higher education continues its transition toward universal access, teaching quality assurance in universities has emerged as a critical concern for scholars and policymakers alike. This case study examines undergraduate teaching quality at a provincial university using fuzzy comprehensive evaluation. Overall, the university's teaching quality received a "good" rating; however, substantial variation was observed across dimensions. Faculty development and teaching resources scored higher, whereas classroom instruction quality and student development support demonstrated comparatively lower ratings. Further analysis identified several key factors constraining teaching quality: underdeveloped mechanisms for faculty professional g
...[truncated 25 chars]

Session Persistence

Medium
Category
Rogue Agent
Confidence
75% confidence
Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · examples/education_examples.md (reported line 75)May include surrounding context.

md
With the in-depth advancement of the popularization process of higher education, the issue of teaching quality assurance in colleges and universities has increasingly become the focus of academic and policy makers' attention. This paper takes a provincial university as a case, adopting fuzzy comprehensive evaluation method to conduct a systematic evaluation of its undergraduate teaching quality. The research found that the overall teaching quality of this university is at a "good" level, but there are significant differences in different dimensions. Specifically, faculty team construction and teaching condition guarantee scored higher, while classroom teaching quality and student development support scored relatively lower. Further analysis showed that key factors affecting teaching quality include: imperfect teacher teaching ability development mechanism, curriculum setting disconnected from market demand, teaching evaluation system too single, etc. Based on the above findings, this paper suggests that colleges and universities should establish a normalized teaching quality monitoring mechanism, improve the teacher teaching ability development support system, and actively promote student-centered teaching model reform.

### ✅ Native English
As higher education continues its transition toward universal access, teaching quality assurance in universities has emerged as a critical concern for scholars and policymakers alike. This case study examines undergraduate teaching quality at a provincial university using fuzzy comprehensive evaluation. Overall, the university's teaching quality received a "good" rating; however, substantial variation was observed across dimensions. Faculty development and teaching resources scored higher, whereas classroom instruction quality and student development support demonstrated comparatively lower ratings. Further analysis identified several key factors constraining teaching quality: underdeveloped mechanisms for faculty professional g
...[truncated 25 chars]

Ae4

Medium
Category
analysis-evasion
Confidence
80% confidence
Finding

Suspicious Unicode normalization or mixed-script content

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

This markdown guide is primarily written in Chinese and presents its guidance as the default format for the skill, while only providing a brief English subtitle rather than an explicit language option. Under the policy, forcing a specific language without user opt-in is a natural-language policy violation unless the locale constraint is clearly documented and justified.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

This markdown file is written entirely in Chinese and presents prescriptive guidance for translation usage without any indication that users may choose another language or locale. Under the policy rule for natural-language violations, forcing a specific language without user opt-in can be a locale-policy issue unless the restriction is explicitly documented and justified.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The file repeatedly instructs the model to translate Chinese academic content specifically into English, starting here with '翻译成适合…发表的英文'. Under the policy rule for all file types, this is a language/locale constraint and the file does not offer an opt-in choice or explain that the skill is intentionally English-only for a documented regional or compliance reason.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The manifest describes a skill for converting Chinese academic papers into publication-ready English via translation and polishing. However, Templates 4 and 5 instruct the AI to draft reviewer responses and submission cover letters, which are adjacent scholarly writing tasks but not part of translating Chinese academic content to English.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The manifest describes a skill for converting Chinese academic papers into publication-ready English and polishing manuscript text. This file instead provides guidance for writing reviewer response letters, which is a related but distinct author-support task not covered by translating or polishing paper content itself.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

This markdown file is entirely framed as a Chinese academic terminology translation database and its instructions are written only in Chinese, effectively imposing a specific language/locale context. Under the policy, language constraints should either provide user opt-in/choice or be clearly justified as region-specific; no such opt-in or justification is present in the file.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
88% confidence
Finding

This skill invites users to submit unpublished academic manuscripts for translation/polishing but does not warn about confidentiality, sensitive data, embargoed results, reviewer-response content, or institutional/IP restrictions. Users may paste proprietary research, patient-related text, or pre-publication findings into an external tool without understanding retention, sharing, or policy risks.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
79% confidence
Finding

This markdown file is entirely framed around producing 'Native English' engineering translations and does not mention any user opt-in or alternative language/locale choices. Under the policy rule, forcing a specific language can be a natural-language policy concern unless the constraint is explicitly justified or optional.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
80% confidence
Finding

This markdown file consistently frames the desired output as 'Native English' and provides no indication that other target languages or locale preferences are supported. Under the language/locale policy rule, a skill should not impose a specific language without user opt-in unless the constraint is clearly documented and justified.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
79% confidence
Finding

The document frames the skill around producing 'good academic English' and related English-writing standards, and similar English-only positioning recurs elsewhere in the file. Because the file does not mention any user language choice or opt-in, this can be read as a locale/language constraint embedded in the skill's design.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
88% confidence
Finding

This markdown file contains natural-language instructional content and therefore falls under SQP-3. The document defaults to Chinese for most explanations while mixing in English examples, but it does not state that the skill is intended specifically for Chinese-speaking users or offer an opt-in choice of language, which can be a language/locale policy concern.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
93% confidence
Finding

This markdown file is natural-language content, so policy checks apply. The document presents all guidance exclusively in Chinese and does not indicate that the language is optional, user-selected, or required for a region-specific audience, which can conflict with a language/locale choice policy.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
92% confidence
Finding

This markdown file presents nearly all instructional content in Chinese, with only a brief English subtitle, which effectively forces a specific language for users. The policy requires either offering a language/locale choice or clearly documenting a justified locale constraint, neither of which appears in the file.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
85% confidence
Finding

This markdown file contains natural-language guidance that presents English-only templates as the default output format, despite the surrounding instruction text being in Chinese. Under the language/locale policy rule, forcing a specific language without user opt-in can be a policy concern when no alternative or opt-in is provided.

Content

No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.