The Agent Skills Directory

[SAFE]: The skill follows standard administrative procedures for setting up a business profile. It defines clear approval boundaries, distinguishing between safe drafting tasks and sensitive actions like changing accounting settings or filing taxes.
[PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection by instructing the agent to ingest data from external sources, such as onboarding documents, accounting exports, and emails (Step 1, Step 5). However, this behavior is central to the skill's functional purpose, and no active exploitation patterns were found.
[DATA_EXFILTRATION]: While the skill prompts for sensitive business information (e.g., entity type, bank account lists, and tax jurisdictions), it instructs the agent to store this data within the user's workspace for reuse. No unauthorized network operations or hardcoded external exfiltration endpoints were detected.

bookkeeping-setup