Agents
VigilQA includes a growing set of AI agents that span generation, review, analysis, and operational monitoring. The runtime experience now covers the full public agent catalog, including requirements extraction and data-quality review flows.
Generation agents
Generation agents run on demand when you trigger the test generation pipeline. They run once per domain or scenario and produce the test artifacts that power your test suite.
| Agent | Stage | Output |
|---|---|---|
| AppExplorer | 1 — Exploration | DOM selector maps per page/endpoint |
| DocGen | 2 — Documentation | Structured test case markdown documents |
| DocReview | 3 — Document QA | Approved docs → ScriptGen; rejected → Approvals |
| ScriptGen | 4 — Script writing | Executable pytest test scripts |
| ScriptReview | 5 — Script QA | Validated scripts → Approvals queue |
Post-run agents
Post-run agents are triggered automatically after every test run. They operate without any manual intervention and produce actionable outputs — alerts, reports, and healing patches. The platform also supports review-oriented agents for requirements extraction and data-quality validation before or alongside generation work.
| Agent | Trigger | Primary output |
|---|---|---|
| ResultAnalyzer | After every run | Failure classification: Bug / Script / Flaky / Environment |
| RegressionGuard | After every run | Regression alert with suspect commit/version |
| CoverageGap | After every run | Coverage % per domain, untested scenarios list |
| FlakyDetector | When ResultAnalyzer flags Flaky | Stability score (0–1), locator instability flag |
| LocatorHealer | On locator failure or FlakyDetector signal | Healing patch queued in Approvals with confidence score |
Monitoring agent activity
All agent activity — both generation runs and post-run analysis — is visible in Agent Activities in the dashboard. Each agent run shows:
- Agent name, trigger, and status
- Start time, duration, and completion time
- Input (which run, which domain, which scenarios)
- Output summary (what was produced, how many items, any errors)
- Link to resulting approvals or reports