Trust
DataEx › Agents › Trust (since 2026-11-01 a sub-tab of Agents, not a rail row; ?tab=harness opens it) is the harness findings alone: what TIDE found about the
platform and what it can do about each finding. Autonomy and Access moved to
DataEx › Agents, and Judgements to the judge’s own page.
GET /api/harness joins the platform’s self-assessment (/api/measures) and the per-control
compliance findings (/api/compliance) into one list. Each finding offers only actions TIDE can
really take:
| Action | What happens |
|---|---|
dispatch_agent | a repository_dispatch to the named agent’s workflow, with the console’s GitHub App token |
file_issue | a tracking issue in the agents repository carrying the evidence |
needs_human | an acknowledgement, recorded |
An agent that addresses a domain but that the console cannot start is shown as an unavailable action with the reason, never hidden.
Acting on a finding (POST /api/harness/{id}/act) runs in this order:
- a session, and
CONSOLE_AGENT_ADMINS; - the finding must be live now;
- the autonomy level must not be
off; - a repeat within the idempotency window replays the earlier result instead of acting twice;
- audit before act: if the intent record is not durable in the audit chain, nothing happens and the answer is 503;
- act, then record the outcome.
auto autonomy level is recorded, but the console
runs no unattended executor.The open findings, worst first, are also the Needs attention section of the Overview.