Menú
How to return a strict scenario result JSON with evidence and no extra prose.
Instalar con Codex o Claude Copia este prompt, pégalo en Codex, Claude u otro asistente, y deja que revise la página de la skill y la instale por ti.
Basado en la clasificación ocupacional SOC
How to consume a GitHub issue body as an agent task prompt. The 9-section template, the lifecycle, the failure modes.
The 12 design principles this codebase is built on. Read first, before any task.
Spec-driven development — the practice of writing the spec before the code. With the 9-section template and the three pnpm spec-check gates.
Differences and deterministic behavior of the mock ADT server used by adt-bench smoke runs.
Test-driven loop for ABAP — inspect, change, check syntax, run unit tests, diagnose, fix.
Standard ABAP object workflow (discover, create or edit, check, activate, validate).
| name | abap-result-contract |
| description | How to return a strict scenario result JSON with evidence and no extra prose. |
Your final answer must be a single JSON object matching this shape:
{
"scenario_id": "string",
"status": "pass | fail | partial",
"summary": "string",
"evidence": [
{ "kind": "object | source | activation | syntax_check | test | diagnostic | tool_result | other", "value": "string" }
],
"changed_objects": ["string"],
"errors": ["string"]
}
Rules:
summary should be 1–3 sentences, plain text, no markdown.evidence should be ordered roughly chronologically. Prefer the strongest evidence (activation, unit test) first.changed_objects is a list of ABAP object names you created or modified.errors is a list of human-readable error strings; empty if nothing failed.status rules:
pass — the task is fully done and validated.partial — you made progress but the task is not complete; explain in summary.fail — you could not do the task; explain in errors.