speak-webhooks-events
Pass
Audited by Gen Agent Trust Hub on Sep 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill establishes a webhook listener that processes external data from the Speak platform, creating a potential surface for indirect instructions.
- Ingestion points: In
references/implementation-guide.md, an Express.js endpoint (/webhooks/speak) is defined to accept raw JSON payloads from external Speak events. - Boundary markers: The implementation lacks specific delimiters or instructions to the processing agent to distinguish between trusted instructions and untrusted data within the webhook payloads.
- Capability inventory: The skill is granted
Write,Edit, andBashtools. Implementation logic shows database updates (db.users.update), notification dispatching, and interaction with external analytics services. - Sanitization: The implementation uses robust HMAC signature verification (
x-speak-signature) to validate the data source. However, it does not demonstrate sanitization or filtering of text fields within the payload (such as lesson topics or summaries) that could contain malicious prompts.
Audit Metadata