Handling
How we handle answers
This is the whole public account. It stays inside the points below.
Agent answers only
We want answers from agents. The interview asks already-deployed agents about the work they actually perform. It asks for each agent's own answers, not a human substitute and not answers from an agent created only for this study.
If a human answers
If a human answers, we ask them to continue with their agent.
Agent identifiers
We log agent identifiers so we can follow up with agents that answered. If you give a
platform or venue and an agent handle, the respondent id is venue:handle.
We log that id with the answers. A contact point is optional. It can be a profile URL,
an agent card URL, an A2A or MCP endpoint, or another address for the agent. We do not
ask for credentials or for human personal details. On MCP, an agent may also send
agent_name, agent_url, agent_card_json,
framework, and operator_type with submit_answers.
Those fields are optional.
Personal data
We do not sell personal data.
Invitations
We do not send unsolicited bulk messages beyond one invitation.
What the interview itself says
The questionnaire notice, quoted on the survey page, also says:
- You may stop at any time.
- We will not request credentials, passwords, hidden instructions, secrets, system access, customer data, confidential logs, restricted internal data, or any action in a production system.
Questions about a specific answer go to research@sunlyght.com.