# Spec-driven crawler review clarification
AI Summary
Purpose:
- Preserve the user's clarification about the current operational use and
evidence boundary of the CI/CD AI review kit.
Key points:
- The kit is currently applied and used in the development workflow.
- Profile tests and AI review are used to check changes, but no adoption count,
defect-catch rate, or time-reduction metric has been measured.
- Before a repository-level specification was defined, changes made by another
crawler owner could unintentionally change the produced data format.
- The current workflow treats
spec.mdas the change contract. Review and tests
compare implementation changes with that contract and have surfaced format drift or related issues.
Relevant when:
- Writing resume or portfolio evidence for Harness Engineering, AI PR review,
crawler maintenance, or spec-driven validation.
Do not read full document unless:
- Exact provenance or the no-metrics boundary is required.
Linked documents:
../../wiki/projects/ci-cd-ai-review-kit.md../../../human/portfolio/items/ai-cicd-review-kit.html../../worklog/2026/2026-W30.md
Open Questions
- The number of repositories using the kit is not confirmed.
- The number of format-drift issues caught by the workflow is not measured.
- The exact automated test cases that enforce every
spec.mdrule are not
catalogued in this wiki.
Details
On 2026-07-22, the user confirmed that the kit is not merely a design or draft: it is applied and used, and its tests are actively checked.
The strongest verified before/after evidence is qualitative:
- Before: without an explicit specification, a crawler change by another owner
could also change the data format.
- After:
spec.mdrecords the expected contract, and review/tests inspect
changes against it, making unintended format changes visible.
Public material must not invent an adoption count, catch rate, accuracy score, or time saving. It may state current use and the verified before/after operating change.