spark-recipe-multi-account-review
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DATA_EXPOSURE]: The skill facilitates access to sensitive email account lists, folder structures, and message metadata through the
sparkCLI tool. This behavior is consistent with the skill's primary stated purpose of providing a multi-account status summary. - [INDIRECT_PROMPT_INJECTION]: The skill processes external data by reading email headers and categories via
spark emails. Malicious content within emails (such as subjects or categories) could attempt to influence agent behavior through indirect prompt injection. - Ingestion points: Data enters the agent context via the output of
spark emailsandspark folderscommands as described inSKILL.md. - Boundary markers: The instructions do not define specific delimiters or instructions to ignore embedded content within the email summaries.
- Capability inventory: The skill uses
spark accounts,spark folders, andspark emailsfor read-only status checks; no file-writing or generic network capabilities are utilized. - Sanitization: No explicit sanitization or filtering of email subject lines or categories is mentioned before presenting the summary.
Audit Metadata