Problem
job run --wait returns status: success, but the agent has no way to verify the actual outcome:
- Were the expected rows loaded?
- Did the schema change unexpectedly?
- Is the row count within normal range?
The agent has to manually chain storage table-detail + row count checks + schema comparison.
Proposal
kbagent job verify --project prod --job-id 456
Output:
- Row counts (loaded vs expected based on recent history)
- Schema diff (if columns changed)
- Data freshness check
- Anomaly flags (e.g., "0 rows loaded but last 7 runs averaged 1,200")
Use case
Critical for agents that run pipelines autonomously — they need confidence that "success" means "actually worked correctly," not just "didn't crash."
Context
Discussion from Devil's Advocate analysis of kbagent's agentic capabilities.
Problem
job run --waitreturnsstatus: success, but the agent has no way to verify the actual outcome:The agent has to manually chain
storage table-detail+ row count checks + schema comparison.Proposal
Output:
Use case
Critical for agents that run pipelines autonomously — they need confidence that "success" means "actually worked correctly," not just "didn't crash."
Context
Discussion from Devil's Advocate analysis of kbagent's agentic capabilities.