The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructs the agent to execute a Python script at marketing-skill/skills/content-production/scripts/brand_voice_analyzer.py. This creates a dependency on an external script not contained within the skill's own package, which could lead to unauthorized code execution if the path or the target script is compromised.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from product descriptions and brand materials to generate code.
Ingestion points: Product details in Step 1 and brand content in Step 2 of the workflow.
Boundary markers: Absent. There are no instructions to the agent to delimit or ignore potential malicious instructions embedded within the user-provided brand content.
Capability inventory: The skill has the capability to execute Python scripts (Step 2) and write file content (Step 7).
Sanitization: There is no explicit validation or sanitization of the input content before it is processed by the agent or the analysis script.
[METADATA_POISONING]: There is a discrepancy between the stated author (seaworld008) and the source URL provided in the metadata, which points to a repository under a different user profile (alirezarezvani). This inconsistency can be used to mask the true origin of the skill.
[DYNAMIC_EXECUTION]: The generation patterns for FAQ sections recommend using dangerouslySetInnerHTML to inject JSON-LD schema markup. If the input data is not properly escaped by the agent, this could result in Cross-Site Scripting (XSS) vulnerabilities in the generated landing page.

landing-page-generator