| name | final-report |
| description | Use when preparing, publishing, or reviewing a final Schemator run report. Ensures the report includes raw artifacts, schema JSON, source details, decisions, verification, and push links instead of only a narrative summary. |
Final Report
Use this skill before calling any Schemator run report complete, especially when publishing it to a scratch repo, PR, issue, handoff doc, or user-facing summary.
Required Contents
A complete final report must include:
- Run identity: source document, run directory or artifact location, timestamp/date, Schemator repo URL, branch, commit, and exact commands used.
- Run outcome: whether the run converged, stable iteration, stopping condition, total iterations, and whether any independent-from-scratch runs were compared.
- Raw schema artifacts: the initial extracted schema graph JSON and the final stable schema graph JSON. Do not only link or summarize them if the user asked for a complete report.
- Human-readable schema summary: initial vs final field/model counts, additions, removals, renames, moves, merges, and unchanged important fields.
- Decision trace: applied changes, skipped proposals, rejected/manual structural proposals, consistency warnings, and the model/field/path affected.
- Context artifacts: project/task context, prompt/default prompt/skill prompt used for decisions, and checksums if available.
- Diff artifacts: machine or human-readable initial-vs-final graph diff.
- Verification: commands run, tests/checks, report generation command, diff generation command, commit SHA, branch, push result, GitHub link, and known gaps.
Completeness Rules
- Do not call a report complete if the raw initial and final schema JSON are missing.
- Do not replace raw artifacts with prose. Prose explains the artifacts; it does not substitute for them.
- Do not report only the final state. Include enough initial state and diff detail for a reader to audit what changed.
- Do not hide skipped or rejected proposals. They are part of the decision record.
