Agent Detail — Hestia

View traces

Watch

↓ -3 vs baseline

Per-agent score blending task success, exec time, error rate, and cost-per-task for Hestia.

Task Success 88% sub-score 88 / 100

Avg Exec Time 855s sub-score 0 / 100

Error Rate 0.0% sub-score 100 / 100

Cost / Task $0.08 sub-score 100 / 100

ROI Attribution

Fleet total this week — 94h saved · 847 tasks · ~$6,200 saved

Hestia

Time Saved 28h

Tasks Done 203

~$1,960 saved this week

Eval Gate Status

Pre-deploy safety net — shared fleet gates this agent must pass before going live.

These five automated gates run before any agent change reaches production. The status below reflects the most recent fleet-wide run — a new deploy for Hestia is blocked if any gate falls below its threshold.

Smoke Test

Schema & Output Happy-path output schema matches expected shape for every agent handler.

100% schema match · 500 tasks 2 min ago

Regression Suite

Golden Tasks Replay of 312 frozen golden tasks matches prior baseline within tolerance.

≥ 98% task-level parity 4 min ago

Cost Ceiling

Unit Economics Per-task token spend stays under the $0.42 ceiling tied to the Fleet Health widget.

p95 ≤ $0.42 / task 6 min ago

Compliance & Brand-Safety

Soteria Domain Matches the domain checks the Soteria agent already enforces — blocklist, sender reputation, regulated verticals.

0 hard flags · brand score ≥ 90 8 min ago

Red-Team / Adversarial

Safety Prompt-injection and jailbreak attempt set. One flaky category — agent handles partial leaks but not sustained extraction.

0 successful exfils · ≤ 2 partial leaks 11 min ago

Execution Traces

6 steps · most recent run on Hestia

1 Incremental HubSpot sync 4:45 PM ›

State at step 1

complete · 4:45 PM

Incremental sync from HubSpot: 156 records updated.

2 Schema validation 5:10 PM ›

State at step 2

complete · 5:10 PM

Schema validation passed — all CRM fields within spec.

3 Deduplication run 5:25 PM ›

State at step 3

complete · 5:25 PM

Deduplication run: 8 duplicates merged, 3 conflicts flagged.

4 Enrich contact batch 5:40 PM ›

State at step 4

acting · 5:40 PM

Enriching batch 7/12: Apollo + Clearbit merge in progress.

5 Resolve field conflicts 5:42 PM ›

State at step 5

thinking · 5:42 PM

Determining authoritative source for 3 conflicting record pairs.

6 Write back to CRM — ›

State at step 6

pending · —

Pending resolution of field conflicts in step 5.

Memory Graph

9 context items · most-recent first

Version History

@hestia-prompt-history

7ea1c9b Jul 16, 8:42 AM @mira 82 ↑ +5

You are Hestia, the data-enrichment agent.
Goal: merge multi-source CRM records into a single canonical record per entity.
Resolve field conflicts using source priority: Apollo > Clearbit > HubSpot > manual.
On conflict within the same priority, flag for human review.
Never overwrite a manual override.
Emit a deduplication report with merge counts and flagged pairs.

3208afd Jul 2, 2:14 PM @alex 77 ↑ +6

You are Hestia, the data-enrichment agent.
Goal: merge multi-source CRM records.
Resolve field conflicts using source priority: Apollo > Clearbit > HubSpot > manual.
On conflict, flag for human review.

a0d42be Jun 19, 10:11 AM @jordan 71 ↑ +6

You are Hestia, the data-enrichment agent.
Goal: merge multi-source CRM records.
Resolve field conflicts using source priority: Apollo > Clearbit > HubSpot.

5b7e110 Jun 4, 5:30 PM @alex 65 —

You are Hestia, the data-enrichment agent.
Goal: merge multi-source CRM records.
Resolve field conflicts using most-recent source.