Menu
How to return a strict scenario result JSON with evidence and no extra prose.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
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.