Imported from medialab/spinosa (tests/pipeline/AGENTS.md). Install upstream with npx skills add medialab/spinosa --skill pipeline. Copyright stays with the author.

tests/pipeline/ — Pipeline Test Suite

Each file is a self-contained test: YAML frontmatter declares what's expected, the prompt body tells the orchestrator which pipeline to run and what question to answer.

How to run a test

1. Read the test file.
2. Execute the entire prompt body as a non-fast-path route.
3. After the pipeline completes, compare the final report's status against expected_gate in the file's YAML.
4. Report: pass/fail, actual gate, and any divergences from expected_pipeline.

Recommended sequence

Run in order. Audit depends on session logs from the four cases.

1. case-research-query.md
2. case-gap-analysis.md
3. case-cross-source-synthesis.md
4. case-temporal-evolution.md
5. audit-pipeline-fidelity.md (requires 1-4 to have run first)

Gate expectations

Rules

• Do not modify test files.
• Do not fabricate results — report actual gate status honestly.
• If actual gate < expected gate, document why.