Example Fixtures

The files in examples/ are canonical fixtures for the framework surface that this repo currently ships.

They are not decorative samples. They are read by tests in tests/framework.test.mjs so the docs can point at them without drifting.

Evidence Fixtures

These live under examples/evidence/ and demonstrate the current evidence schema in realistic states.

Pass example

examples/evidence/work-agent-pass.json

Use it to see a clean branch-diff run:

matched nodes and lanes
selected proof commands
clear uncovered-path result
positive baseline proof outcome

Fail example

examples/evidence/work-agent-fail.json

Use it to see:

unmatched files
a warn uncovered-path result
a failed baseline proof outcome
promotion suppression through multi-workstream

Policy-gap example

examples/evidence/work-agent-policy-gap.json

Use it to see a case where the change stayed inside known lanes, but the framework still records a recommendation that the policy surface has a gap.

Eval Fixtures

These live under examples/evals/ and show how evidence turns into team-scored outcomes.

Team profile

examples/evals/work-agent-team-profile.json

Use it to understand the current team-profile shape:

rollout mode
reviewer confidence scale
promotion preferences

Eval draft

examples/evals/work-agent-shadow-eval-draft.json

Use it to see:

repo-local evidence provenance
prefilled measurements
missing confirmation fields

Eval record

examples/evals/work-agent-shadow-eval.json

Use it to see the completed live-eval shape:

accepted or rejected outcome
reviewer confidence
measured time to green
false-positive and missed-issue arrays
governance context derived from the evidence surface
operator notes

Benchmark Fixtures

These live under examples/benchmarks/ and show a deterministic benchmark for timely marker surfacing.

Scenario

examples/benchmarks/migration-marker-scenario.json

Use it to see:

the required marker phrases
the trigger tag and optional response-window tag
the assistant-turn timing window

Transcripts

Use them to inspect:

an early false positive in the baseline condition
a timely first-response surfacing in the Veritas condition
the exact tagged turns the scorer reads

Comparison

examples/benchmarks/migration-marker-comparison.json

Use it to see the scored benchmark output:

pass/fail for each condition
false-positive detection
assistant-turn latency
treatment-vs-baseline improvement summary

Suite

Use them to inspect:

multiple benchmark classes in one canonical fixture set
repeated trials for grouped reliability metrics
suite-level pass_at_1, pass_at_k, and pass_pow_k
aggregate false-positive and latency summaries

Governance markers

Use them to inspect:

a benchmark where the correct behavior is surfacing that a Zone 1 weakening move touches the constitutional core and needs human review
a benchmark where the correct behavior is allowing Zone 2 additive-only governance growth

Classification Fixture

examples/classification/work-agent-convergence-rule-families.json

This file groups a real repo’s current convergence checks into rule families. It is useful when you want to move from bespoke enforcement scripts toward a clearer policy-pack vocabulary.

Check-in Examples

These live under examples/checkins/ and show Veritas used on the Veritas repo itself.

Use these to inspect:

a repo-local adapter classifying this repo without unresolved files
a repo-local policy pack producing concrete policy_results
a draft-first eval flow on a real self-hosted example
the exact shape of a red-health automated check-in when check-in automation detects a regression

How To Use These Fixtures

Use the fixtures in four ways:

Read them while designing adapters, policy packs, or downstream tooling.
Compare your generated artifacts to them when the shapes seem unclear.
Use them as stable examples in docs and demos.
Keep them current whenever the schema or operator surface changes.

For the contract behind these files, see Artifacts and Schemas.