Skip to main content

Report Template (Locked, Chat + File)

Every format-drift-scanner delivery follows this shape. Chat-first compact summary, then optional file write to Downloads. Never deviate from the section order; never rename headers.

Two Output Modes​

Mode A: Compact chat summary (default). Rendered every run. Lightweight; safe to scan in a Studio session.

Mode B: Full report file (opt-in). Written to Downloads folder only when user confirms. Uses the full section-by-section shape.

Mode A: Compact Chat Summary Format​

**Format Drift Scanner: {workbook_name}**
Placements: {N} measures · {N} dates · {N} text objects · Verdict: {Ready with warnings | Ready to publish | Blocking issues surface}

Findings: {N} Blocking · {N} Warning · {N} Info{ · {N} Unscanned if > 0}

Top items:
1. **[F-{RULE-ID}]** {finding_summary} → {short Desktop fix}
2. **[F-{RULE-ID}]** {finding_summary} → {short Desktop fix}
3. **[F-{RULE-ID}]** {finding_summary} → {short Desktop fix}

Next: {one-line next step}

Want the full report? I can save it to your Downloads folder as `.md` (default) or `.txt`. Say the word.

Compact summary rules.

  • 8-line ceiling in chat before the offer line.
  • Verdict is exactly one of Ready to publish, Ready with warnings, Blocking issues surface.
  • Top items ordered Blocking first, then Warning, then Info.
  • If total findings is zero, render "Findings: 0 · Ready to publish" and skip the Top items block.
  • Unscanned count appears in the counts line only when > 0.
  • Offer line is exactly: Want the full report? I can save it to your Downloads folder as `.md` (default) or `.txt`. Say the word.

Mode B: Full Report File Format​

Used only when the user opts into a file write. Path: {Downloads}/{workbook_name}_format-drift.md (or .txt). Fresh-filename discipline: increment _v2, _v3 on collision. Path resolves OS-appropriately: %USERPROFILE%\Downloads\ on Windows, $HOME/Downloads/ on macOS. Never hard-code a tilde-expanded path.

# Format Drift Scanner: {workbook_name}

**Placements audited:** {N} measure placements · {N} date pills · {N} text objects
**Data sources:** {ds_names_list}
**Verdict:** {Ready to publish | Ready with warnings | Blocking issues surface}
**Report generated:** {ISO-8601 timestamp}

## Findings by Severity

### Blocking ({N})
- **[F-{RULE-ID}]** {finding_summary}
- **Scope:** {within_ds ds_name | cross_ds | workbook}
- **Element:** {field_name or element_category}
- **Differing format strings:** {baseline} vs {alternates_list}
- **Fix in Desktop:** {desktop_fix_instruction}

### Warning ({N})
{Same shape as Blocking. Empty section renders as _None._}

### Info ({N})
{Same shape. Empty renders as _None._}

## Findings by Drift Type

| Drift Type | Count |
|---|---|
| Currency (F-CURRENCY-01) | {N} |
| Decimal (F-DECIMAL-02) | {N} |
| Date (F-DATE-03) | {N} |
| Separator (F-SEPARATOR-04) | {N} |
| Font Family (F-FONT-FAMILY-05) | {N} |
| Font Size (F-FONT-SIZE-06) | {N} |
| Negative Treatment (F-NEGATIVE-07) | {N} |

## Baselines Per Drift Type Per DS

{Rendered as a list. Baselines are the first-seen format string per HC#11.}

- **{drift_type}** for **{field_name}** in **{ds_name}**: `{baseline_format_string}`

## Unscanned Placements ({N})

{Rendered only if count > 0. Otherwise omit this section.}

- **{element_type}** on **{worksheet_name}** — `reason: {tool_failure | user_skip | fallback_unavailable}`
- **Notes:** {short_string}

## Next Steps

{One or two-line pointer. Never generic inspiration.}

Section Rendering Rules (Mode B, Full File Only)​

Header line. Always # Format Drift Scanner: {workbook_name}. Placements audited, data sources, verdict, and timestamp on the four lines under the header.

Verdict logic:

  • Blocking issues surface if blocking > 0.
  • Ready with warnings if blocking == 0 AND (warning > 0 OR info > 0).
  • Ready to publish if blocking == 0 AND warning == 0 AND info == 0.

Severity sections. Always render Blocking, Warning, Info in that order. Empty sections render as _None._. Never omit; empty is a signal of a clean check.

Findings by Drift Type table. Always renders. Zeros are legitimate values; do not omit rows.

Baselines per Drift Type per DS. Always renders. Includes baselines even when no drift finding fired for that drift type; downstream diff-mode reuse depends on the baseline being recorded.

Unscanned Placements section. Only renders when count > 0. Omit entirely when count is zero.

Next Steps line. Concrete pointer: "Publish-readiness handoff emitted" if clean, or "Fix Blocking Currency drift first, then re-run" if issues surface. Never generic advice.

Overflow Handling​

  • 20+ findings. Group Info findings under a <details> block after the first 5 per drift type. Blocking and Warning always fully rendered.
  • Multiple DSes. Order findings by DS name alphabetically for stable diffs across runs.
  • Very wide workbook (100+ placements). Add a per-worksheet appendix at the end of Mode B if any single worksheet has 10+ placements audited.

File-Write Confirmation Flow​

  1. Mode A chat summary renders.
  2. Offer line appears at the end.
  3. If the user opts in ("yes," "save it," "put it in Downloads," ".md," ".txt"):
    • Confirm format choice if not already stated (default .md).
    • Resolve the Downloads folder OS-agnostically. On Windows: %USERPROFILE%\Downloads\. On macOS: $HOME/Downloads/. Never hard-code a tilde-expanded path or a drive letter; render the fully-resolved absolute path.
    • Confirm the exact save path before writing.
    • Write the file per Mode B shape.
    • Report the path back in chat.
  4. If the user declines or doesn't respond, no file is written. Handoff schema still emits to session context regardless.

Never write a file silently. HC#1 read-only discipline extends to Downloads-folder writes. Always confirm the path before saving.

Fresh-filename discipline. If the target path already exists, append _v2 (then _v3, etc.) before the extension.

Naming convention. The _format-drift suffix is reserved to this skill; never reuse it for other skills' exports.

Refuse to Fabricate​

Never render a finding without a rule-cited baseline and at least one differing format string. Never invent a drift rule ID. Never render a Blocking finding without confirming the drift is within a single DS (cross-DS is Info-by-default per HC#12). Empty sections are signals of a clean workbook, not gaps to paper over.