The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill requires a tool named treg to function. If the tool is not present, it provides instructions for the user to visit https://treg.to/llms.txt to manually install and sign in to the service. This introduces a dependency on unverified software from a non-trusted domain.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from multiple external sources, including target websites, search engine results, advertisements, and social media platforms (YouTube, TikTok, Instagram, X, LinkedIn, Reddit) to generate growth playbooks and content reports.
Ingestion points: Data is ingested throughout the analysis process, specifically in Steps 1 (site mapping), 4 (search and AI results), 5 (advertisements), 6 (social mentions), 7 (social media content and image analysis), and 8 (backlinks and product loops) using tools like treg.web.extract and various third-party scraping APIs.
Boundary markers: While the skill enforces a 'SEEN' vs 'INFERRED' logic for the agent's reasoning, the instructions do not specify explicit delimiters or 'ignore embedded instructions' warnings for the raw text and captions retrieved from external sites.
Capability inventory: The skill has the capability to write multiple files to the filesystem, including playbook.md, playbook.html, content.md, content.html, and several CSV files (e.g., moves-vs-growth.csv, posts.csv).
Sanitization: The instructions do not define any sanitization, escaping, or filtering protocols for the external content before it is interpolated into the generated reports and playbooks.

rea-growth