consultsciences/sakirea-skills

Checks that every factual claim in a draft maps to a cited source that actually supports it, then emits a gaps table of unsupported, mis-attributed, and uncited claims. Use for manuscript, report, or memo review before submission, evidence audits, reference-to-claim reconciliation, and tightening over-claimed prose.

Ranks clinical sources by evidence level, extracts a claim-to-evidence table from a mixed source pile, and returns a structured triage report. Use when a clinical question needs rapid appraisal of study quality, contradictory trials, guideline-vs-trial conflicts, or an evidence-graded summary before deeper review.

Verifies that every number, unit, and annotation in a table or figure caption can be traced to a provenance record or is explicitly labeled TODO, then returns a traceability matrix of matched, orphaned, and unresolved entries. Use for manuscript figure QA, table audits, dashboard number checks, and pre-submission provenance passes.

Builds reproducible PubMed E-utilities queries (term construction, field tags, date filters), runs esearch plus esummary with plain Node fetch, and records exact provenance for every search. Use for PubMed or MEDLINE lookups, literature sweeps, update alerts, documented search strings, or any task needing a repeatable citation search.

Monitors key-free regulatory feeds (Federal Register API, EU clinical-trial registry and CT.gov record patterns) and turns new documents into a dated watch list with provenance. Use for FDA or agency rulemaking watches, trial-registry change monitoring, regulatory horizon scanning, and repeatable daily or weekly feed checks.

Chooses an appropriate baseline statistical test, computes mean, sample SD, standard error and 95% confidence interval for a numeric sample with a stdlib-only script, and formats effect-size-first reporting. Use for small-sample summaries, test selection questions, CI-over-p-value requests, or checking how a result should be reported.