scenario-authorlisted

# Skill — Scenario author ## When this loads Anything that writes or edits a test scenario: `scenario_create`, `scenario_update`, `scenario_set_steps`, `cod-doc scenario new`, or a request phrased as "describe how we'd test X", "add scenarios for capability Y". ## What a scenario is here A scenario is a **claim about what should be true**, written down so a human or an agent can act on it. It is not a test, and it does not know whether a test exists. RFC 24 splits the problem in two, and this skill only touches one half: | Half | Who owns it | Where it lives | |---|---|---| | Intention — the claim | you, via `scenario_*` | `scenario` tables, projected to `docs/system/scenarios/` | | Evidence — does a test prove it | the structure producer in ai-reviewer | `scenario_assessment` (STR-002), not yet built | ## The five kinds (RFC 24 §9, fixed vocabulary) Do not invent a sixth. The producer joins assessments on these exact names. | Kind | Use it for | Example from this repo | |---|---|---| | `happy_path` | the capability doing its job with valid input | completing a task recomputes plan progress | | `error_path` | a named failure the system must handle, not crash on | `task_update_status` from `todo` without `via_checkout` is refused (ADO-039) | | `boundary_value` | the edge of a range, an empty set, the first/last element | exporting a group whose scenarios were all retired | | `invariant` | something that must hold across every operation | every mutating service writes