The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill defines a workflow where the agent generates code (landing pages, tools, dashboards) based on natural language descriptions provided by users. This establishes a surface for indirect prompt injection if those descriptions contain instructions intended to influence the agent's behavior.
Ingestion points: Natural language descriptions provided by users to generate 'GTM assets' in SKILL.md.
Boundary markers: None explicitly defined to separate user-provided content from agent instructions during the generation phase.
Capability inventory: The skill references tools like Claude Code, Cursor, and v0 which possess file-system and code-execution capabilities.
Sanitization: Not present for input descriptions.
[COMMAND_EXECUTION]: The skill includes scripts/check-output.py, which is designed to validate text deliverables. Technical analysis confirms the script is benign; it reads a user-specified local file and performs string length and keyword checks without any network operations or unsafe execution patterns.
[SAFE]: No evidence of credential harvesting, unauthorized external downloads, or hidden persistence mechanisms was found. The external platforms referenced (v0, Lovable, Bolt.new, Cursor, Replit) are well-known technology services. The skill metadata and documentation are consistent with the stated purpose of assisting in AI-powered development.

vibe-coding