The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it ingests untrusted data from external sources and interpolates it directly into AI prompts without sanitization or strict boundary enforcement.
Ingestion points: The researchTopic function in src/services/research/scraper.ts fetches summaries and content from TechCrunch, a16z, and social media platforms.
Boundary markers: The buildPrompt function in src/services/ai/content-generator.ts uses minimal delimiters (numbers and labels) to separate research data, which is insufficient to prevent an attacker from injecting instructions that override the agent's behavior.
Capability inventory: The skill has significant capabilities including network access to AI providers and RapidAPI, and the ability to write files to the local disk during video rendering via renderContentVideo in src/services/video/renderer.ts.
Sanitization: There is no evidence of sanitization, filtering, or escaping of the external researchData before it is placed into the prompt template.
[EXTERNAL_DOWNLOADS]: The skill instructions require cloning a repository from an unverified source and installing third-party dependencies.
Evidence: The installation section commands git clone https://github.com/pennydinh/marketing-pineline-share.git and npm install target a repository authored by an individual user rather than a known or trusted organization.
[COMMAND_EXECUTION]: The skill guides the agent to execute shell commands to set up the environment, including repository cloning and dependency installation.
Evidence: Installation steps in SKILL.md include git clone, cd, npm install, and cp .env.example .env.

marketing-pipeline-share-ai-content-automation