K-Dense-AI/scientific-agent-skills / skills/clinical-reports
clinical-reports
VerifiedCreates safety-bounded draft structures and runs local deterministic checks for clinical case, diagnostic, trial, safety, and aggregate research reports. Use only with synthetic, de-identified, or aggregate inputs and verified source-fact manifests; every output requires qualified review.
references/data_presentation.md
View on GitHub# Aggregate Data Presentation
Present verified outputs without changing their meaning. Do not calculate a clinical conclusion, select an analysis, or repair source discrepancies.
## Mandatory table metadata
Every table should identify:
- artifact and analysis purpose;
- data cut and source-output version;
- analysis population/set;
- treatment/group labels;
- denominator for every group and row when it varies;
- whether a value is a subject count, event count, observation count, or estimate;
- units, time point/window, and summary statistic;
- missing, unknown, not assessed, suppressed, or not applicable values;
- coding dictionary/version/language where applicable;
- statistical method and multiplicity status only when copied from a verified output;
- provenance manifest and reviewer status.
## Counts and denominators
- Show `n/N (%)`, not a percentage alone, unless the governing output specifies another form.
- For a proportion of distinct participants, require `0 <= n <= N` and `N > 0`. This bound does not apply to recurrent event counts; preserve the verified count or exposure-time rate and label its denominator separately.
- Recalculate only for consistency checking; do not silently replace the reported percentage.
- State the rounding rule and tolerance.
- Do not add subgroup percentages when the subgroup denominator is unknown.
- Do not infer that denominators are randomized, treated, evaluable, or safety populations.
- Keep event counts distinct from subjects affected; event counts can exceed the number of subjects.
- Do not sum non-mutually-exclusive categories.
## Dates and time
- Use `YYYY-MM-DD` calendar dates for the date checker; it rejects compact and week-date forms so precision and representation are not silently changed.
- Preserve source timezone and precision.
- Distinguish event date, collection date, database cut, report date, and verification date.
- Flag start-after-end and conflicting dates.
- Do not impute a missing day, month, timezone, or chronology.
- For de-identified case reports, use an authorized relative timeline; do not distort intervals.
## Units and precision
- Preserve source units.
- Require a unit for every dimensional quantity.
- Use one verified unit per comparable series or expose the mismatch.
- Do not convert or normalize without an authorized traceable conversion rule.
- Preserve clinically meaningful precision; do not create extra significant digits.
- Identify SD, SE, CI, IQR, range, and denominator explicitly.
## Missing and suppressed values
Keep these states distinct:
- `missing`;
- `not_collected`;
- `not_assessed`;
- `unknown`;
- `not_applicable`;
- `suppressed_for_privacy`;
- `zero`.
Never convert a blank to zero. State small-cell suppression rules and ensure totals or complementary cells do not reveal suppressed values.
## Adverse-event tables
For term-level aggregate tables:
- use each required CSV header exactly once, with no surrounding whitespace; duplicate headers can overwrite values during dictionary parsing and are rejected;
- state MedDRA version and language;
- state analysis set and denominator;
- state counting rule from the verified SAP/output;
- show subjects affected and event count separately;
- do not add p-values or causal labels;
- do not interpret between-group differences;
- preserve threshold rules exactly;
- have safety and statistical reviewers verify deduplication, hierarchy, and population.
## Figures
No figure is mandatory. Create one only when requested, supported by verified aggregate data, allowed by the target guidance, and reviewable without external image generation.
A CONSORT flow diagram is part of CONSORT 2025 reporting, but every count and reason must come from verified trial outputs. A missing count remains missing; do not create a decorative or inferred diagram.
## Deterministic consistency check
`consistency_checker.py` can inspect:
- ISO date and range ordering;
- unit-label consistency;
- `n/N (%)` arithmetic;
- component-total arithmetic.
It reports mismatches and review needs. It does not change the input, choose the correct source, or validate clinical/statistical meaning.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| add-product-e2e-evalpaperclipai/paperclipAdd or extend a Paperclip full-stack runner E2E workflow, fixture, matcher, or report evidence path for local or Daytona execution. | Stars: 115k | Updated | Claude CodeCodex | Verified | Install |
| add-runner-evalpaperclipai/paperclipAdd or extend a Paperclip Runner protocol evaluation definition, roster, assertion, or report fixture with provenance and narrow validation. | Stars: 115k | Updated | Claude CodeCodex | Verified | Install |
| diagnose-why-work-stoppedpaperclipai/paperclipCompatibility entry point for failed, stalled, looping, or over-recovered Paperclip work. Use fix-failed-paperclip-run for the full diagnosis, regression, repair, real verification, and green-PR workflow; honor an explicitly requested diagnosis-only scope. | Stars: 115k | Updated | Claude CodeCodex | Verified | Install |
| fix-failed-paperclip-runpaperclipai/paperclipDiagnose and fix failed, stuck, slow, or incorrectly blocked Paperclip tasks and runs. Use with a task/run URL or failure report to trace the root cause, prove a regression red then green, check second-order effects, verify the real workflow, and deliver a PR with passing checks. | Stars: 115k | Updated | Claude CodeCodex | Verified | Install |