Skip to main content

ingestion-validation

Selects and runs evidence-based validation for Python ingestion, chunking, jobs, Azure AI Search contracts, deployment assets, and the operator frontend. Use whenever repository behavior or engineering assets change.

Source facts

Repository
Azure/agent-app-ingestion
Last source activity
July 25, 2026 at 19:40
Detected SKILL.md language
English
Stars
188
Forks
101

Install options

The review-first prompt is selected by default. You can switch to a direct command or download a local copy.

Review the source files

Read SKILL.md and any companion files shown by SkillsMP before deciding whether to install.

Showing SKILL.md

SKILL.md
Source instructions · Read-only preview
name
ingestion-validation
description
Selects and runs evidence-based validation for Python ingestion, chunking, jobs, Azure AI Search contracts, deployment assets, and the operator frontend. Use whenever repository behavior or engineering assets change.
# Ingestion validation Start with the narrowest existing command that covers the change. ## Engineering-agent assets ```text python -m pip install --requirement .github/scripts/requirements.txt python .github/scripts/validate-agentic-assets.py ``` ## Python Run focused tests first: ```text python -m pytest tests/<affected_test>.py ``` Then run the maintained suite when risk or repository policy requires it: ```text python -m pytest ``` Do not require live Azure credentials for unit tests. Reuse existing fakes, stubs, and golden fixtures. ## Frontend From `frontend/`, use the scripts defined in `package.json`: ```text npm test npm run lint npm run build ``` ## Runtime integration For changes to formats, sources, jobs, embedding, Search payloads, or deployment: 1. Build and run the existing container or use the applicable `scripts/deploy.ps1` or `scripts/deploy.sh` path. 2. Exercise a representative source and document. 3. Verify the terminal job result and expected Azure AI Search documents, fields, ACL metadata, and deletion behavior. 4. Inspect logs and correlated audit events without copying document content, credentials, or private environment names into public artifacts. If a command is unavailable or live validation is unsafe, report the missing dependency and residual risk. Do not substitute an unrelated passing check.
View on GitHub