Back to skill

Security audit

Christa 资深塑形教练

Security checks for vulnerabilities and agentic risk

Overview

This is a coherent fitness and nutrition coaching skill, but users should treat its health advice as general guidance and be careful with generated HTML reports.

Install only if you want a Chinese-oriented fitness/body-shaping coach. Do not use it for diagnosis, pregnancy, chronic disease, medication, eating disorder, kidney disease, diabetes, or injury decisions without a licensed professional. If generating HTML reports, ensure user-provided text is escaped or rendered as plain text before opening the file in a browser.

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
references/report-template.md:34
Finding

Unescaped User-Controlled Content in Generated HTML Report

Content
View full analysis
{{summary}}
① 目标清晰度{{s1}}/20
{{e1}}
② 身体与基础状态{{s2}}/20
{{e2}}
③ 饮食管理能力{{s3}}/20
{{e3}}
④ 运动与塑形能力{{s4}}/20
{{e4}}
⑤ 生活方式与执行力{{s5}}/20
{{e5}}

接下来最重要的 3 件事

① {{t1}}
② {{t2}}
③ {{t3}}
教练寄语 {{coach}}
``` The corresponding instructions require these placeholders to be replaced with assessment-derived content, but do not require HTML escaping or sanitization: ```text Replace {{placeholders}} with real content. ` ...[truncated 3190 chars]
Remediation
View remediation
` to `>` - `"` to `"` - `'` to `'` 2. **Treat assessment-derived content as plain text.** Do not preserve HTML supplied by users in `summary`, dimension explanations, tasks, or the coach message. 3. **Use context-specific validation.** - Scores must be integers from `0` through `20`. - Percentages must be numeric values from `0` through `100`. - SVG coordinates must match a strict numeric-coordinate format and should be calculated internally. - Do not place arbitrary user-controlled strings into style, SVG, or other attribute contexts. 4. **Avoid direct string-based HTML construction where possible.** Build the report through a trusted HTML templating system with automatic escaping, or assign dynamic values through DOM `textContent` rather than `innerHTML`. 5. **Sanitize any intentionally supported formatting.** If limited markup such as highlighting must be permitted, generate that markup internally. Otherwise, process content with an allowlist-based HTML sanitizer that rejects scripts, event-handler attributes, unsafe URLs, embedded objects, and SVG active content. 6. **Add a restrictive Content Security Policy to the generated report**, for example: ```html ``` 7. **Add adversarial test cases** containing closing tags, event handlers, script elements, malformed attributes, SVG payloads, and encoded markup. Verify that each payload is rendered visibly as text and never interpreted as DOM content. ]]>
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (9)

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 144)May include surrounding context.

md
| `assets/avatar.png` | 教练头像 512×512 |

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The skill description is unusually broad and includes many common health, diet, and fitness topics such as protein, carbs, blood sugar, fasting, keto, and supplements. This increases the chance the skill will be invoked in situations where the user intended a narrower or different capability, causing inappropriate routing and potentially unqualified advice in sensitive health-adjacent contexts.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The SOP explicitly requires coaches to present information in a fixed professional tone and to avoid disclosing that the interaction follows a system workflow or backend scoring process. This removes user choice over communication style and obscures that scoring is being applied, which can undermine transparency and informed consent in a health-adjacent coaching context.

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 explicitly instructs the agent to answer strictly according to the provided coaching system, and the content is entirely in Chinese, creating a strong default bias toward Chinese responses without any user language preference check. This can cause the skill to ignore or override the user's chosen language, reducing usability and potentially misleading users who cannot read the output.

Content

No source excerpt is available for this finding.

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.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

This knowledge file gives extensive prescriptive advice on fasting, ketosis, supplements, electrolyte dosing, endocrinology, and condition-adjacent interventions while explicitly instructing the agent to answer strictly from this material. In a consumer coaching skill, the absence of medical disclaimers, contraindication handling, and referral thresholds can cause the model to deliver unsafe health guidance to users with diabetes, kidney disease, eating disorders, pregnancy, medication interactions, or other conditions.

Content

No source excerpt is available for this finding.

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

The document title, content requirements, and HTML template all prescribe Chinese-language output, including lang="zh-CN" and Chinese-only report text. This is a natural-language locale policy constraint with no opt-in, user choice, or explanation that the skill is region-specific.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
88% confidence
Finding

The file content and instruction framing require answers to follow this source as written, and the source is entirely in Chinese without offering an alternative language or user choice. SQP-3 applies to all file types and flags language-policy issues when a skill effectively forces a specific language or locale without explicit opt-in or documented justification.

Content

No source excerpt is available for this finding.