The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided article content and themes to generate search queries, allowing for potential indirect prompt injection if the inputs contain malicious instructions.
Ingestion points: The skill ingests the "slug ou caminho do artigo" and the "tema central do artigo" as specified in SKILL.md.
Boundary markers: Absent. There are no explicit instructions or delimiters used to ensure the agent ignores instructions embedded within the processed article data.
Capability inventory: The skill possesses the ability to perform network requests (search and download photos via API), write files to the local file system (saving images and credit notes), and modify existing files (updating the article frontmatter).
Sanitization: Absent. While REFERENCIA.md provides strategies to extract "concrete objects" from the text, this is a logical guideline rather than a programmatic security control or sanitization filter.
[EXTERNAL_DOWNLOADS]: The skill is designed to download photographic assets from external stock photo providers.
Evidence: SKILL.md instructs the agent to "Busque fotos reais... selecione a foto... baixe e corte".
Context: The downloads target well-known services (such as Unsplash) and are essential for the primary purpose of the skill.
[COMMAND_EXECUTION]: The skill requires the agent to perform file system operations and image processing tasks.
Evidence: The flow requires downloading and cropping images ("baixe e corte nas dimensões") and updating file metadata (frontmatter).

inboundfy-especialista-blog-imagem