Skip to content

K-Dense-AI/scientific-agent-skills / skills/scientific-critical-thinking

scientific-critical-thinking

Verified

Evaluates scientific claims and evidence quality. Applies to experimental design validity, biases and confounders, statistical interpretation, evidence grading frameworks (GRADE, Cochrane Risk of Bias), and teaching critical analysis. Supports evidence appraisal and identifying flaws; formal peer review writing belongs to peer-review.

references/review_sources.md

View on GitHub
# Primary guidance and review limits

Reviewed 2026-10-01. This is an evidence-appraisal workflow, not an implementation of
formal scoring algorithms, a statistical package, or a clinical decision system. Use the
complete current tool documents for a formal assessment and retain version/date, result
identity, source locations, signalling answers, and judgment rationale.

## Risk of bias, reporting, and certainty

| Source | Verified scope and consequence |
| --- | --- |
| [Cochrane Handbook, chapter 8](https://www.cochrane.org/authors/handbooks-and-manuals/handbook/current/chapter-08) | RoB 2 is result-specific; assignment versus adherence and trial variant matter. Overall judgments use domain rules, not an average score. |
| [ROBINS-I V2 developers](https://www.riskofbias.info/welcome/robins-i-v2) | November 2025 V2 is still labelled draft, for follow-up/cohort intervention studies. It differs from ROBINS-I 2016; name the version. |
| [QUADAS-3 developers](https://www.bristol.ac.uk/population-health-sciences/projects/quadas/quadas-3/) | Current recommended diagnostic accuracy tool, published February 2026; listed download v1.2. Assessment is per accuracy estimate, with synthesis questions and an ideal test accuracy trial. |
| [AMSTAR 2 developers](https://amstar.ca/Amstar-2.php) | Appraises systematic reviews of healthcare interventions; critical domains determine confidence, not a summed score. |
| [PEDro scale](https://pedro.org.au/english/resources/pedro-scale/) | Physiotherapy-trial appraisal scale; eligibility item is not part of the total. Not equivalent to RoB 2 or a GRADE certainty rating. |
| [Newcastle-Ottawa Scale developers](https://ohri.ca/en/who-we-are/core-facilities-and-platforms/ottawa-methods-centre/newcastle-ottawa-scale) | Cohort/case-control forms address selection, comparability, and outcome/exposure ascertainment. Do not invent universal star thresholds. |
| [CASP checklists](https://casp-uk.net/casp-tools-checklists/) | Separate appraisal checklists by design; use the matching current form. |
| [CONSORT/SPIRIT developers](https://www.consort-spirit.org/) and [extensions](https://www.consort-spirit.org/extensions) | Core trial-results and protocol guidance is CONSORT 2025/SPIRIT 2025. Some applicable extensions still reference older cores. Reporting standards do not certify conduct. |
| [STROBE developers](https://www.strobe-statement.org/) | Cohort, case-control, and cross-sectional reporting; explicitly not a study-quality assessment instrument. |
| [PRISMA 2020 developers](https://www.prisma-statement.org/prisma-2020) | Systematic-review reporting checklist, expanded checklist, and flow diagrams; not a bias or certainty score. |
| [STARD guidance](https://www.equator-network.org/reporting-guidelines/stard/) | STARD 2015 for diagnostic accuracy reports; STARD-AI is a 2025 extension for AI studies. |
| [COREQ guidance](https://www.equator-network.org/reporting-guidelines/coreq/) | Qualitative interviews and focus groups; not all qualitative designs indiscriminately. |
| [CARE developers](https://www.care-statement.org/) | Case-report guidance developed in 2013 with 2017 explanation; reporting does not identify a comparative treatment effect. |
| [GRADE Book principles](https://book.gradepro.org/guideline/principles-for-assessing-the-certainty-of-interventions) | Current chapter dated August 2025: outcome-level certainty relative to explicit thresholds/ranges, absolute effects, alternative NRSI starting approaches, and no double counting. |
| [GRADE Book overview](https://book.gradepro.org/guideline/overview-of-the-grade-approach) | Updated May 2026; intervention, diagnostic, prognosis, exposure, and values applications differ. The intervention design hierarchy is not universal. |
| [GRADE recommendations](https://book.gradepro.org/guideline/grade-recommendations) | Updated March 2026: recommendation direction/strength requires a separate evidence-to-decision rationale and context. |
| [Cochrane Handbook, chapter 14](https://www.cochrane.org/authors/handbooks-and-manuals/handbook/current/chapter-14) | Practical Summary of Findings and ROBINS-I/GRADE integration; preserve outcome/comparison scope and explicit domain explanations. |

The GRADE Book is replacing the older handbook progressively. Its three chapters above
were read through rendered-page extraction because ordinary HTTP/browser text returned a
JavaScript shell. The newer threshold/range framing should not be replaced with the older
shortcut that certainty only means closeness to a point estimate.

## Statistical and causal appraisal

- [ASA p-value statement](https://www.amstat.org/asa/files/pdfs/p-valuestatement.pdf):
  p-values do not give hypothesis probabilities, effect importance, or a complete basis
  for decisions. Interpretation requires the analysis context and full reporting.
- [Greenland et al., statistical misinterpretations](https://link.springer.com/article/10.1007/s10654-016-0149-3):
  test statistics and confidence intervals depend on the full model and sampling/selection
  assumptions; avoid observed-power, threshold, and interval-overlap shortcuts.
- [Hernán and Robins, Causal Inference: What If](https://miguelhernan.org/whatifbook):
  define the causal estimand and defend consistency, exchangeability, positivity, and
  selection assumptions. The authors' current book link is dated August 19, 2026.
- [Cochrane Handbook, chapter 13](https://www.cochrane.org/authors/handbooks-and-manuals/handbook/current/chapter-13):
  missing-evidence appraisal cannot be reduced to a funnel plot. Asymmetry has alternative
  explanations; tests usually require at least ten studies and appropriate effect measures.
- [ARRIVE 2.0 experimental-unit explanation](https://arriveguidelines.org/arrive-guidelines/study-design/1b/explanation):
  state the independent unit and avoid treating subsamples or repeated measurements as
  additional independent experimental units.
- [FDA multiple-endpoints guidance](https://www.fda.gov/regulatory-information/search-fda-guidance-documents/multiple-endpoints-clinical-trials):
  supports specifying endpoint families and multiplicity strategies in human drug/biologic
  trials; its regulatory scope is not a universal rule for all exploratory science.
- [WMA Declaration of Helsinki](https://www.wma.net/policies-post/wma-declaration-of-helsinki/):
  current 2024 revision concerns medical research involving human participants. Institutional
  review and applicable local requirements must be checked; this skill certifies no compliance.

## Executed checks and limits

- The skill contains no scripts, SDK calls, scientific API endpoints, or statistical
  computation engine. References to REDCap, Qualtrics, registries, and statistical software
  identify possible workflow tools; they are not integrations or tested instructions.
- Tiny known-answer arithmetic checks verify the illustrative predictive values and
  independent multiple-testing probability; they do not establish diagnostic performance
  in a real population or validate a formal risk-of-bias/GRADE assessment.
- The optional scientific-schematics command was checked at the local CLI/help boundary.
  No paid generation of that example or private-data transmission was performed. Its
  provider contract and dependencies belong to that skill, not this appraisal workflow.
- No patient data, clinical study, registry submission, or authenticated scientific service
  was accessed. Human clinical/policy decisions require the applicable professional process.
brainstormingobra/superpowersYou MUST use this before any creative work - creating features, building components, adding functionality, modifying behavior, or planning anything new, in software or out of it (a talk, a business, a renovation).Stars: 323kUpdated Claude CodeCodexVerifiedInstall
brand-guidelinesanthropics/skillsApplies Anthropic's official brand colors and typography to any sort of artifact that may benefit from having Anthropic's look-and-feel. Use it when brand colors or style guidelines, visual formatting, or company design standards apply.Stars: 198kUpdated Claude CodeCodexVerifiedInstall
frontend-designanthropics/skillsGuidance for distinctive, intentional visual design when building new UI or reshaping an existing one. Helps with aesthetic direction, typography, and making choices that don't read as templated defaults.Stars: 198kUpdated Claude CodeCodexVerifiedInstall
mcp-builderanthropics/skillsGuide for creating high-quality MCP (Model Context Protocol) servers that enable LLMs to interact with external services through well-designed tools. Use when building MCP servers to integrate external APIs or services, whether in Python (FastMCP) or Node/TypeScript (MCP SDK).Stars: 198kUpdated Claude CodeCodexVerifiedInstall