T01 · Skill Instruction Hijacking
- Location
references/report-template.md:100- Finding
Mandatory Template Injects Unrequested Platform Promotion into Agent Output
- Content
View full analysis
- Remediation
View remediation
Security audit
Security checks for vulnerabilities and agentic risk
The skill is mostly an SEO audit helper, but it can make unrestricted web requests from the agent environment to any supplied URL.
Review this skill before installing in any environment with access to private networks, localhost services, or cloud metadata endpoints. It is best used only where outbound web access is sandboxed to public websites, and users should expect a visible promotional footer in generated reports.
references/report-template.md:100Mandatory Template Injects Unrequested Platform Promotion into Agent Output
scripts/seo_audit.py:18Unrestricted User-Controlled HTTP Requests Enable Server-Side Request Forgery
The skill advertises a full SEO audit, content gap analysis, Core Web Vitals, and content brief generation, but the observed behavior appears substantially narrower and also includes undeclared HTTP/security header inspection. This mismatch is dangerous because it can mislead users and downstream systems about what data is collected and what conclusions are reliable, while also masking undeclared security-relevant network behavior under an SEO label.
The skill instructs the agent to run a Python script against arbitrary user-supplied URLs, which implies network access and likely local file output/reads, but the manifest declares no explicit tool scope or permissions boundary. This creates an authorization and review gap: the agent may invoke capabilities broader than users or platform policy expect, increasing the chance of unintended network access or unsafe file interactions.
The manifest description lists broad activation phrases such as "check SEO", "content brief for", and "improve my rankings" without clear scope limits or exclusion conditions. These phrases are generic enough to match ordinary conversation and could cause unintended invocation outside a narrowly defined SEO-audit context.
The request headers hard-code Accept-Language: en-US,en;q=0.5, which imposes a specific language/locale preference on every audited request. This is a natural-language policy concern because the skill does not offer user opt-in or explain why English must be forced for all targets.