dial-your-context
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTIONOBFUSCATION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to use
curlfor making HTTP requests to the Sanity API. This is used to fetch initial context, query schemas, and test GROQ filters. - [DATA_EXFILTRATION]: The skill requires a
SANITY_ORGANIZATION_TOKENto be sent in theAuthorizationheader toapi.sanity.io. This is a vendor-owned domain (sanity-io) and the token is necessary for the legitimate operation of configuring the vendor's service. - [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external Sanity datasets to determine how the agent should interpret specific fields. This presents an attack surface where malicious content in the dataset could influence the generated instructions. The skill mitigates this through strict workflow rules requiring the agent to verify all claims with the user and provide supporting evidence for every instruction created.
- [OBFUSCATION]: Employs URL encoding for the
instructionsandgroqFilterquery parameters. This is a standard requirement for passing complex strings in HTTP GET/POST requests and is not used here to mask malicious behavior.
Audit Metadata