The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill uses tools like tavily/search-assistant and exa/search to ingest content from the public internet for newsletter curation. This creates a surface for indirect prompt injection where malicious instructions embedded in external web pages could influence the agent's behavior.
Ingestion points: Commands utilizing belt app run tavily/search-assistant and belt app run exa/search in the 'Content Sourcing' section of SKILL.md fetch untrusted data into the agent context.
Boundary markers: Absent. The instructions do not provide delimiters or warnings to the agent to ignore potentially malicious instructions found within the curated search results.
Capability inventory: The skill allows access to the Bash tool (via the belt CLI), which could be leveraged if the agent is manipulated by injected content.
Sanitization: Absent. There are no guidelines for the agent to filter or escape the content fetched from external sources before using it to generate newsletter sections or social media teasers.

newsletter-curation