pipa-huddle-beta

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.90). The skill explicitly tells the agent to capture a generated session id and reuse/print it (e.g., --session <id> and the session URL), which requires embedding that identifier verbatim in commands and outputs and thus poses an exfiltration risk if the id is sensitive.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). The runtime path forwards the user’s spoken/text turns from the hosted voice page at https://voice.usepipa.com/s/<session-id> through the relay to the local OpenCode bridge (POST /api/turn) for each turn, so outsider-authored free text (the user’s dictated content) is directly ingested by the LLM.