viant/endly/endly-dsunit-batch-validation
Claude Code and agent skill by viant/endly
- community
Checking that the SKILL.md file is still in the repository.
Validate Endly batches by use-case IDs across database expectations, asynchronous logs, and aggregate outputs with bounded waits.
Install
npx vskill@latest install viant/endly/endly-dsunit-batch-validationSKILL.md
Collect expected outcomes per case while executing the case, carrying a stable TagId/correlation ID and the selected fixture IDs. Append expected records to a batch collection rather than overwriting another case's expectations. Keep payload expectations separate from status-only assertions; a successful request can still produce incorrect persisted rows or logs.
Start validator/log:listen and reset its observation buffer before the batch. Configure format, file mask and correlation extraction so records map to the correct use case. After request generation, validate the accumulated expectations with validator/log:assert using bounded LogWaitTimeMs/LogWaitRetryCount. Preserve per-case descriptions and IDs in failures. Avoid turning a long arbitrary sleep into evidence of asynchronous completion; use the harness's retry/readiness checks.
For persisted/aggregate outcomes, use dsunit:expect or query with Expect after writes have settled. Inspect the live checkPolicy and matching options before changing them; never weaken a matching policy just to accept unrelated rows. Validate both expected presence and unwanted extras when the contract requires it. Preserve mappings, dictionary prerequisites and per-case hydration throughout.
Focused reruns must narrow both executed cases and accumulated expectations; otherwise a batch validator can report errors for cases that were not executed. Use actual discovered TagIDs and inspect assertion counts. Operation status is failed when assertions fail; zero assertions or a skipped case is not a pass. Retrieve operation_events pages filtered to assertion/failure types and inspect only the state paths needed to diagnose the mismatched case.
When the listener can fall back to FIFO before a correlation ID is indexed, use a case-specific inclusion rule or queue so it cannot consume an unrelated earlier record. Preserve that rule in a focused run; a longer wait alone does not repair ambiguous matching.
Dispatch database checks with endly_dsunit_expect/query/compare and correlated log checks with endly_validator_log_listen/reset/assert. Each action returns an operation; check status and assertion counts before continuing.
Security
- Security scanPassed, 90 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingskills/endly-dsunit-batch-validation/SKILL.md
Popularity
- 7-day trend score
- 5
Refreshed 11 Oct 2026, 19 days ago.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/viant/endly/endly-dsunit-batch-validation)