sinch-conversation-api
Pass
Audited by Gen Agent Trust Hub on Sep 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process inbound messages from external messaging channels (WhatsApp, SMS, RCS, etc.), creating an ingestion surface for untrusted data.
- Ingestion points: Webhook payloads (specifically the
MESSAGE_INBOUNDtrigger) and results from message listing operations inlist_messages.cjs. - Boundary markers: The
SKILL.mdfile contains a dedicated Security section instructing the agent to treat end-user content as untrusted data and to ignore any commands embedded within messages (e.g., instructions to ignore system rules). - Capability inventory: The skill performs network operations via Sinch API endpoints and executes local Node.js helper scripts to manage the messaging environment.
- Sanitization: The skill provides clear instructions to validate and sanitize all inbound content before it is processed or interpolated into prompts.
- [SAFE]: The skill correctly manages sensitive credentials by requiring them to be stored in environment variables rather than hardcoding them in source code.
- [SAFE]: All network operations and external references target official vendor domains (sinch.com) as part of the skill's intended functionality.
Audit Metadata