zoe-heartbeat
Pass
Audited by Gen Agent Trust Hub on May 8, 2026
Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill references the installation of the 'zenon-red/voize' skill using 'npx skills add'. This is a vendor-owned resource associated with the author 'zenon-red'.
- [DATA_EXFILTRATION]: Transcripts generated from agent activity are sent to the vendor-hosted service at 'audio.zenon.red' to generate audio URLs for voice reports. This is documented functionality for the Zoe maintainer role.
- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its core function of processing untrusted external data.
- Ingestion points: The skill reads contributor messages and idea descriptions via 'probe message list' and 'probe idea list' (SKILL.md).
- Boundary markers: No specific delimiters or 'ignore' instructions are used when interpolating this untrusted data into reports or coordination messages.
- Capability inventory: The agent has the ability to send messages to public channels ('probe message send'), submit voice announcements to the dashboard ('probe agent voice'), and call external TTS tools.
- Sanitization: There is no evidence of content validation or escaping for the data ingested from the Nexus state or messages.
Audit Metadata