Back to skill

Security audit

Mouthpiece

Security checks for vulnerabilities and agentic risk

Overview

This is a plain markdown workplace-advice skill, but it asks users to build and automatically retain a sensitive workplace and financial profile while making privacy promises it cannot enforce.

Install only if you are comfortable sharing workplace context with your agent. Avoid providing employer-identifying details, debt or financial-pressure information, or sensitive dispute facts unless necessary, and check whether your agent host actually stores profiles locally and lets you inspect or delete them.

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)

T09 · Insecure Skill Coding Practices

Warning
Location
SKILL.md:23
Finding

Unnecessary Sensitive Profiling and Unsupported Privacy Guarantees

Content
View full analysis

Vulnerability Details

File Location: SKILL.md:23-27, SKILL.md:75-77, user-profile.md:36-44, user-profile.md:75-84, user-profile.md:90-107
Vulnerability Type: Excessive personal-data collection and unverifiable privacy claims
Risk Level: Medium

Relevant excerpts translated into English from the source:

markdown
### First use
Learn about the user's workplace background to provide precise advice:
- Organization type
- Position level
- Financial situation, including mortgage obligations, living paycheck to paycheck,
  or inherited wealth
markdown
- Responses are adjusted according to workplace background, position, and financial pressure.
- When the background changes, tell the Agent to update the profile.
- Conversations are confidential and used only for personalized advice.
markdown
## History

### Previously discussed scenarios
- Automatically records the types of scenarios discussed by the user.

### Preferred response style
- Automatically learns from the responses selected by the user.

### Feedback history
- Records user feedback to optimize future advice.
markdown
The profile is updated automatically, and subsequent advice is adapted to the new situation.

### Data privacy
- Profile information is used only to provide personalized advice.
- All information is stored locally.
- Conversation content is not shared or uploaded.

Technical Analysis

The Skill solicits financial circumstances and employment information even though these details are not necessary to produce general workplace communication templates. It also describes automatic maintenance of consultation history, preferences, and feedback.

The project contains only Markdown instructions. It provides no implementation of consent controls, data minimization, encryption, access restrictions, retention periods, secure deletion, or storage isolation. It also contai ...[truncated 1888 chars]

Remediation
View remediation

Remediation Suggestions

  1. Remove financial-status questions unless a concrete and proportionate need is documented.
  2. Make all profiling optional and provide useful generic advice when users decline.
  3. Obtain explicit, informed consent before retaining profile fields, consultation history, preferences, or feedback.
  4. Default to session-only processing rather than persistent storage.
  5. Clearly identify every collected field, its purpose, storage location, retention period, access boundary, and deletion procedure.
  6. Replace absolute claims such as “all information is stored locally” and “conversation content is not uploaded” with statements verified against the actual Agent runtime and model provider.
  7. Provide commands or documented procedures that allow users to inspect, correct, export, and permanently delete their profile.
  8. If persistence is implemented, apply restrictive file permissions, encryption where appropriate, retention limits, and safeguards preventing profile data from entering unrelated prompts or sessions.
  9. Collect only the minimum context needed for each response and avoid retaining raw workplace-conflict narratives.
Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (12)

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The activation examples are broad enough to match ordinary workplace complaints, which can cause the skill to engage unintentionally in general conversation rather than through a clearly scoped invocation. That increases the chance of unsolicited behavioral steering and accidental collection of contextual personal or workplace information from users who did not realize they were invoking a specialized persona.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill asks for sensitive profiling data, including employment context and financial status, but does not provide a clear privacy warning, minimization rationale, or handling caveat at the point of collection. This creates unnecessary privacy risk because users may disclose sensitive socioeconomic information without understanding why it is needed or how much is actually required for the feature to work.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The skill content is entirely Chinese and frames responses in Chinese without offering user-language detection or choice. This can cause unsafe misunderstandings, reduce user comprehension of risk warnings, and lead to inappropriate outputs when the surrounding agent or user operates in another language.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The when_to_use trigger is broad and subjective ('when the user wants to vent workplace troubles, needs reply suggestions, or wants witty pushback'), which can cause the skill to activate in many adjacent situations without clear boundaries. Overbroad activation increases the chance the agent will steer normal conversations into this persona, collect unnecessary sensitive workplace context, or provide adversarial/combative messaging where a neutral assistant response would be safer.

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 skill presents all instructions and examples exclusively in Chinese, which can amount to forcing a specific language without user opt-in. The file does not indicate that the skill is region-specific or that users may choose another language, so it appears to violate the language/locale policy criterion.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill content is entirely in Chinese and provides no indication that it will adapt to the user's preferred language. In an agent setting, forcing a single language without user opt-in can reduce usability, cause misunderstanding of guidance, and exclude users who cannot safely interpret the output.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The entire skill content is written in Chinese and provides only Chinese response templates, with no indication that users may choose another language or that the skill is intentionally limited to a Chinese-speaking context. Under the policy, language-specific behavior without user opt-in or clear justification is a natural-language policy violation.

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 file presents all instructions, templates, and warnings only in Chinese, which effectively forces a specific language on users. Under the policy rule, language constraints should either be optional or clearly justified; no such opt-in or justification appears here.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
82% confidence
Finding

All user-facing instructions, examples, and response formats are written in Chinese, and the README does not state that the skill is limited to Chinese-speaking users or provide any opt-in language selection. Under the stated policy, forcing a specific language without user choice can be a natural-language policy issue unless the locale 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
91% confidence
Finding

The file presents all instructions, examples, and usage text exclusively in Chinese, which effectively forces a specific language experience. Under the policy, language constraints should either offer user choice or be explicitly documented and justified.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
92% confidence
Finding

The entire skill is written in Chinese and provides no indication that language choice is optional or that the skill is intentionally limited to a Chinese-speaking audience. Under the stated policy, forcing a specific language without opt-in is a natural-language policy violation.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
90% confidence
Finding

The entire template and instructions are written only in Chinese, and the file does not indicate that language choice is optional or that the skill is intentionally limited to a Chinese-speaking context. Under the policy, language constraints should be opt-in or clearly justified when a specific locale is required.

Content

No source excerpt is available for this finding.