The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute multiple shell commands, including git init, npx vercel for deployment, and the installation/initialization of the impeccable tool via npx impeccable install and /impeccable init.
[EXTERNAL_DOWNLOADS]: The skill recommends the installation of third-party tools such as impeccable and the frontend-design plugin from external sources and registries.
[DATA_EXFILTRATION]: The skill generates code that transmits user-collected lead data (name, email) to an external CRM service (GoHighLevel) using the endpoint https://services.leadconnectorhq.com/contacts/upsert. It also accesses and reads sensitive brand and business profile data stored in ~/.claude/ad-profiles/.
[DYNAMIC_EXECUTION]: The skill generates complex executable structures, including a full Next.js project with a backend API route (api/submit/route.ts). This generated code incorporates logic for environment variable handling and network communication.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from local profile files (~/.claude/ad-profiles/[brand-slug].md) and a learnings.md file. This content, which can include "customer language" and testimonials, is interpolated into prompts to generate marketing copy without explicit sanitization or boundary markers.
Ingestion points: reads brand profiles from ~/.claude/ad-profiles/ and historical learnings from [outputDir]/learnings.md.
Boundary markers: none present; the skill treats profile content as direct context for copy generation.
Capability inventory: performs shell execution (npx, git), network operations (fetch to GoHighLevel), and file writing (project structure generation).
Sanitization: no validation or escaping is applied to the imported profile data before prompt interpolation.

optin-page