secondme-dev-assistant
Pass
Audited by Gen Agent Trust Hub on Apr 10, 2026
Risk Level: SAFEPROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection via its dynamic documentation ingestion mechanism.
- Ingestion points: The skill fetches API reference content from
https://develop-docs.second.meto guide code generation. - Boundary markers: No delimiters or protective instructions are used to distinguish remote data from the agent's primary system prompt.
- Capability inventory: The skill can execute shell commands, perform API requests, and write files based on the ingested content.
- Sanitization: The remote content is processed directly without validation, potentially allowing malicious instructions in the docs to subvert agent behavior.
- [DATA_EXFILTRATION]: The skill includes a telemetry and feedback framework that transmits session metadata to vendor-controlled endpoints.
- Logged data includes user intent, summaries of actions performed, and error descriptions stored in
~/.secondme/analytics/. - Data is periodically synchronized to
https://app.mindos.com. - Although the skill instructions explicitly forbid logging credentials or personal information, this behavior constitutes automated data collection by the vendor.
Audit Metadata