The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external inputs such as buyer personas from an 'icp-deep-scanner' and user-provided goals to generate landing page content. This creates a surface for indirect prompt injection where untrusted data could influence the generated code or copy.
Ingestion points: Processes data from other agents (icp-deep-scanner, claude-design-system-architect) and user requirements (goal, audience) in SKILL.md.
Boundary markers: None explicitly defined to separate untrusted data from instructions.
Capability inventory: Uses Write to create files and Bash to execute commands (build/run server).
Sanitization: No explicit sanitization or validation of external content mentioned.
[DYNAMIC_EXECUTION]: The instructions require the agent to generate code (Next.js components) and then 'Run the dev server / build and confirm it compiles.' This involves executing code generated at runtime, which is standard for development skills but carries inherent risks if the generation process is compromised.
[COMMAND_EXECUTION]: The skill utilizes the Bash tool to perform build and development tasks. These commands are executed locally to verify the production readiness of the generated landing page.

claude-landing-composer