Verification
Verification checks acceptance criteria against real captured evidence — command output, in most cases — rather than trusting a task's own report that it's done.
AgentDock Pro
Verification is part of AIgency, available only in AgentDock Pro.
An agent saying it finished a task isn't the same as the task actually being done. Verification closes that gap by evaluating a task's acceptance criteria against evidence AIgency actually captured, rather than the agent's own completion message.
How a criterion gets evaluated
Each acceptance criterion has a required check (commonly a command to run) and a description of what passing looks like. A verification-kind task (see Execution) runs that check and records the real output as evidence, then compares it against the criterion.
Note
For agents with no structured tool-call transport (notably Antigravity), verification evidence is captured on a best-effort basis rather than from a guaranteed structured tool result — see Agent adapter architecture.
Outcomes
A criterion's evaluation lands in one of a fixed set of states: pending, running, passed, failed, blocked, or skipped. A verification report aggregates the results across all of a task's (or run's) criteria and feeds quality gates — which turn that aggregate result into an accept/reject decision — and, when a criterion fails, repair loops.