The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructions direct the user to clone a project repository from a well-known service (github.com/pennydinh/marketing-pineline-share.git) and install dependencies using standard package managers (npm or yarn). While the repository owner (pennydinh) differs from the skill author (aradotso), the provided snippets demonstrate standard Next.js and Remotion application logic without suspicious behaviors.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it is designed to ingest data from untrusted external sources (such as TechCrunch, Twitter, and LinkedIn) and process that content through LLMs to generate marketing copy.
Ingestion points: Research data is fetched from various social and news platforms via the lib/research/scraper.ts module.
Boundary markers: Not present in the provided code snippets; data is interpolated into prompts for content generation.
Capability inventory: The skill leverages AI providers (Anthropic/OpenAI) and local video rendering capabilities (Remotion).
Sanitization: No explicit sanitization or filtering of fetched research content is shown in the prompt interpolation snippets. This is considered a common characteristic of automated research tools.
[CREDENTIALS_UNSAFE]: The skill documentation correctly advises users to manage sensitive API keys (Anthropic, OpenAI, RapidAPI) using a .env.local file, which is a standard security practice for local development to prevent secret exposure.

marketing-pipeline-share-automation