T09 · Insecure Skill Coding Practices
Warning
- Location
prompts/03-reputation-referral.md:31- Finding
Selective Review Solicitation Enables Review Gating
- Content
View full analysis
- Remediation
View remediation
Security audit
Chimney Sweep Fireplace Kit
Security checks for vulnerabilities and agentic risk
Overview
This markdown-only marketing skill does not run code, but some prompt templates can generate misleading public credential, rating, and review-request claims.
Install only if you will manually review and edit outputs before publishing. Verify every credential, rating, license, insurance, and jurisdiction-specific claim, and change the review sequence so all customers receive the same neutral opportunity to leave an honest public review.
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 (3)
T09 · Insecure Skill Coding Practices
Warning
- Location
prompts/02-service-pages-chimney-schema.md:31- Finding
Unconditional JSON-LD Template Can Publish False Professional Credentials
- Content
View full analysis
- Remediation
View remediation
T09 · Insecure Skill Coding Practices
Warning
- Location
prompts/04-digital-ads-local-seo.md:31- Finding
Fixed Advertisement Instructions Can Fabricate Certification and Rating Claims
- Content
View full analysis
- Remediation
View remediation
Vulnerability Patterns
- Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
- MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
- Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
- Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
- Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (5)
Intent-Code Divergence
High
- Category
- Not specified by scanner
- Confidence
- 99% confidence
- Finding
The CSIA credential gate correctly blocks 'CSIA Certified' when the business is only 'in training,' but Output B hard-codes the Facebook ad description as '[BUSINESS_NAME] — CSIA Certified.' That contradiction can directly induce false credential claims in public ads, exposing users to deceptive advertising, credential misrepresentation, and possible licensing or trade-association enforcement issues.
- Content
Natural-Language Policy Violations
Medium
- Category
- Not specified by scanner
- Confidence
- 80% confidence
- Finding
All required output copy, headlines, CTAs, and fixed phrases are specified in English, and the prompt does not offer any opt-in for another language or locale. Under the policy, forcing a specific language without user choice can be a natural-language policy violation unless the locale restriction is explicitly justified.
- Content
Intent-Code Divergence
Medium
- Category
- Not specified by scanner
- Confidence
- 98% confidence
- Finding
The prompt’s FTC compliance section bans '5-star' and star-count requests in public-facing content, yet the Google ad headline guidance explicitly asks for '[REVIEW_COUNT] 5-Star Reviews.' This internal contradiction can cause the model to generate non-compliant advertising copy, creating regulatory and consumer-protection risk for downstream users who trust the prompt’s compliance framing.
- Content
Natural-Language Policy Violations
Low
- Category
- Not specified by scanner
- Confidence
- 80% confidence
- Finding
The README frames the skill around US market assumptions and specific Nevada/legal standards, such as US home sales volumes and Nevada contractor scope, while presenting the skill as a general marketing kit. Because the file does not offer a locale choice or clearly scope activation/use to US or Nevada users only, it risks a language/locale policy mismatch for users outside that context.
- Content
Missing User Warnings
Low
- Category
- Not specified by scanner
- Confidence
- 76% confidence
- Finding
The prompt asks for a trust headline using "[REVIEW_COUNT] 5-Star Reviews" even though the compliance section says public-facing content must not use "5-star" requests or star-count language. Because this file is a markdown prompt that instructs generation of outward-facing ad copy, a warning or clarification should be included so users understand this sensitive marketing/compliance constraint before use.
- Content