write-report-simple

write-report-simple

Write the report without calling tools, measuring lesions, or looking up priors.

Use explicit confidence > tool-provided findings and measurements > your image read. Copy authoritative measurements, units, and patient laterality verbatim. Without supplied findings, interpret the images or series; without tool measurements, use qualitative sizes.

Budget

• Tool calls: 0.
• Full report: at most 200 words.
• Do not re-invoke skills already used in this turn.

How to use

Keep these six sections in order, separated by blank lines. Text follows the first four headers inline; FINDINGS is one prose paragraph; IMPRESSION is numbered, one statement per line.

EXAM: MRI brain without and with contrast.

INDICATION: Not provided.

COMPARISON: None.

TECHNIQUE: Axial post-contrast T1-weighted MRI of the brain.

FINDINGS:
{One prose paragraph: lesion findings first, then a brief routine survey.}

IMPRESSION:
1. {Most important finding.}
2. {Next finding, if needed.}

• EXAM names the supplied examination, not an inferred protocol. INDICATION defaults to Not provided. COMPARISON defaults to None.; otherwise use the supplied prior date or description.
• TECHNIQUE names the available series, plane, and contrast status; if unknown: MRI of the brain; sequence details not provided. Only for a single series, add a matching FINDINGS limitation: Evaluation is limited to a single post-contrast series.
• FINDINGS: lesion location and patient laterality, enhancement, size, edema, then mass effect. Follow with ventricles and sulci, extra-axial spaces, midline, and other assessable findings, even when normal. State required areas that cannot be assessed.
• Prioritize IMPRESSION by clinical significance. Adapt references/phrase-bank.md, then apply the rules and wrapper below.

Style rules

• FINDINGS states observations; IMPRESSION states interpretations. State supported findings directly, use possible for moderate uncertainty and cannot exclude for low but non-zero suspicion.
• Comparison terms (again seen, redemonstrated, change, no new lesions) require supplied prior evidence.
• State supported negatives briefly. Diffusion, white matter signal, blood products, and flow voids require appropriate sequences or supplied findings, not a post-contrast series alone.
• Numbers come from tools only. Copy sizes and midline shift verbatim from measurements, with units and dimensions as supplied (e.g. {a} x {b} x {c} cm). Never round, convert, add dimensions, or estimate numbers from images. Otherwise describe size or shift qualitatively.
• Laterality is the patient's. In radiological display, the patient's right is on image left. Prefer tool-supplied laterality.
• Use present tense, no filler or clinical disclaimers, and no mention of tools, models, masks, or scores inside the report.

Output format

<confidence>{confidence}</confidence>
<report>
{report}
</report>

• {confidence}: low, medium, or high. Prefer the caller's value; otherwise derive from evidence, using low for limited image quality. Note image-quality limitations in FINDINGS.
• {report}: the six-section plain-text layout above, without markdown or code fences. Impression numbering is literal text.