The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its core functionality of processing untrusted external data.
Ingestion points: The SKILL.md file specifies the ingestion of keyword exports, customer call transcripts, survey responses, and web search results from forums like Reddit and Quora.
Boundary markers: The instructions lack delimiters or specific directives to the AI to ignore instructions potentially embedded within the ingested external data.
Capability inventory: The skill's primary function is data analysis and document generation; no dangerous subprocess calls, file-write operations, or network exfiltration scripts were identified.
Sanitization: No evidence of input validation or sanitization routines for external content was found.
[NO_CODE]: The skill consists exclusively of Markdown documentation and natural language instructions. No executable scripts (Python, JavaScript, Shell) or binaries are included in the package.
[SAFE]: The skill references several well-known technology services and headless CMS platforms (e.g., Sanity, Contentful, Strapi) for informational and selection purposes. These references are strictly for documentation and do not pose a security risk.

content-strategy