| name | ingestion-pipeline |
| description | Use this skill when working on Faultline fixture intake from public sources, including staging review, duplicate rejection, promotion into fixtures/real, and post-promotion baseline validation. Trigger it for requests about running the ingestion pipeline, ingesting fixtures, reviewing staging candidates, promoting accepted cases, or validating new real-world corpus additions. |
Ingestion Pipeline
This skill is for running Faultline's deterministic fixture intake loop with deliberate source diversity.
Use it when the task involves:
faultline fixtures ingestfaultline fixtures reviewfaultline fixtures promote- validating a promoted fixture against the real-corpus baseline
- building a varied intake batch across multiple public-source adapters
Do not use it for general coding work, broad repository review, or playbook authoring that is not tied to a fixture intake task.
Read First
Workflow
Refer to docs/modules/ingestion-workflow-steps.md for the complete step-by-step workflow.
Key execution notes for agents:
- Follow the modular workflow steps precisely
- Use source adapters (
github-issue, gitlab-issue, stackexchange-question, discourse-topic, reddit-post) from the reference module
- Apply all guardrails during execution
- Document all deliverables as specified
Essential commands include:
./bin/faultline fixtures stats --class real --json (check current corpus mix)./bin/faultline fixtures stats --class real --check-baseline (validate baseline integrity)- Bias toward underrepresented adapters when current corpus statistics show gaps
- Handle repository-local risk findings in
internal/engine/testdata/source/ as source-playbook fixtures
Guardrails
Refer to docs/modules/ingestion-guardrails.md for complete guardrails, quality standards, and acceptance criteria.
Deliverable
Refer to docs/modules/ingestion-deliverables.md for the complete deliverables specification and report structure.
