The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external web sources such as TechCrunch, Twitter, and LinkedIn to inform content generation. This untrusted data is fed directly into Claude or OpenAI models without visible sanitization or boundary markers, creating a surface for indirect prompt injection.
Ingestion points: researchTopic scraping logic in SKILL.md.
Boundary markers: None provided in the documentation.
Capability inventory: renderMedia for video production, CronJob for automation, and npx command execution.
Sanitization: No evidence of input filtering or instruction escaping.
[EXTERNAL_DOWNLOADS]: The installation process requires cloning a repository from an untrusted source (github.com/pennydinh/marketing-pineline-share.git). The repository name contains a typo ('pineline' instead of 'pipeline'), which is an inconsistency compared to the skill's title and purpose.
[PERSISTENCE]: The skill includes instructions for implementing CronJob to automate content generation cycles, which allows for persistent, scheduled execution of the pipeline logic on the host system.
[DYNAMIC_EXECUTION]: The video generation feature utilizes @remotion/bundler and @remotion/renderer to bundle TypeScript components and execute them at runtime to produce media assets.
[COMMAND_EXECUTION]: Installation and usage guides prompt the user to execute several shell commands, including git clone, npm install, and npx remotion, which involve downloading and running code from external sources.

marketing-pipeline-share-content-automation