The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes transaction descriptions retrieved from external sources and user-provided allocation instructions, creating a surface for indirect prompt injection attacks.
Ingestion points: Transaction metadata retrieved via the transaction_search tool and user-supplied strings describing the split allocation in SKILL.md.
Boundary markers: The skill does not define specific delimiters or instructions to treat transaction data as untrusted content.
Capability inventory: The skill uses transaction_search and categorize tools in SKILL.md; it does not have high-risk capabilities like file system modification or network access.
Sanitization: The skill implements logical validation by summing amounts but lacks explicit text sanitization or escaping for transaction descriptions.

smart-split