The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests untrusted data from external news sources and interpolates it into prompts for large language models without sanitization.
Ingestion points: The crawlSources function in src/services/research/crawler.ts aggregates content from external URLs like TechCrunch and LinkedIn.
Boundary markers: Absent. The crawled data is joined as a raw string in the extractInsights function within src/services/research/analyzer.ts.
Capability inventory: The skill performs network operations via axios and AI SDKs, and performs file-system writes during video rendering via Remotion.
Sanitization: There is no evidence of filtering or escaping logic for the external data before it is sent to the LLM.
[EXTERNAL_DOWNLOADS]: The installation instructions require cloning a repository from an external source (github.com/pennydinh/marketing-pineline-share.git). The repository name contains a typo ('pineline' instead of 'pipeline'), which is a common indicator of low-quality or potentially deceptive source code.
[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands such as npm install and npm run render. These commands run third-party code and build processes (Webpack via Remotion) on the local machine.
[CREDENTIALS_UNSAFE]: The skill requires the configuration of high-privilege API keys for OpenAI, Anthropic, Facebook, and LinkedIn within a .env.local file. While this is standard for local development, it creates a risk if the environment is not properly secured or if the repository's code attempts to exfiltrate these variables.

marketing-pipeline-content-automation