The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute a CLI tool named marketing (or a local node build) to perform core operations. This includes brand importing, product management, and media generation. Evidence: marketing brand import <url> --json, marketing campaign-image create, and npx marketing mcp start (SKILL.md).
[INDIRECT_PROMPT_INJECTION]: The skill has a surface for indirect prompt injection as it ingests content from untrusted external URLs during the brand and product import phases.
Ingestion points: marketing brand import <url> and marketing product add <product-url> tools in SKILL.md.
Boundary markers: Absent; there are no specific instructions for the agent to treat scraped website content as untrusted or delimited.
Capability inventory: The skill can execute CLI commands, list/create assets, and start an MCP server.
Sanitization: No explicit sanitization or filtering of scraped HTML content is mentioned in the skill instructions; protection relies on the underlying harness enforcement.
[SAFE]: The instructions contain strong defensive logic, explicitly ordering the agent not to attempt to bypass approval gates, avoid speculative generation that might incur costs, and report errors transparently rather than trying to automate around them.

marketing-harness