structured-analytic-skills plugin
Structured analytic techniques as agent skills — each implements a published method, cites it, and ships deterministic companion tools.
How to install the structured-analytic-skills plugin
Run both commands inside Claude Code. The first adds the marketplace once; the second installs the plugin.
/plugin marketplace add radarist/structured-analytic-skills
/plugin install structured-analytic-skills@structured-analytic-skillsWhat's inside structured-analytic-skills
70 parts read from the plugin's own files.
Skills 70
Claude loads it when the task matches- _TEMPLATE
No description in the file.
- abstain-or-escalate
Decides what to do with a claim that verification could not confirm — drop it, refuse it in plain words, escalate it as a named open question, or report the disagreement between conflicting sources — instead of shipping a hedged guess. Use when verification has already failed: "there is no source for this", "the sources contradict each other", "should this claim be dropped or flagged?", "abstain rather than guess". Not for running the verification pass itself (use `grounded-answer`) or for checking published numbers against live sources (use `grounded-fact-check`).
- amstar2-review-appraisal
Appraises a systematic review or meta-analysis with AMSTAR 2 (AMSTAR-2; Shea et al., BMJ 2017): answers the 16 checklist items from the review's report, flags flaws in the seven critical domains and derives the overall confidence rating — High, Moderate, Low or Critically low — never a summed score. Use when deciding whether a published review can be relied on, or ranking reviews for an umbrella review: "appraise this systematic review with AMSTAR 2", "how trustworthy is this meta-analysis?", "which AMSTAR 2 items are critical?". Not for one trial (use `assess-study-bias`), a body of evidence per outcome (use `evidence-appraisal`) or conducting a review (use `systematic-review`).
- analysis-of-competing-hypotheses
Ranks rival explanations of a question by which one the evidence contradicts least (Heuer's Analysis of Competing Hypotheses — ACH — CIA), then names the single piece of evidence the ranking hinges on. Use when a question has two or more plausible explanations and confirmation bias would be costly — "why is adoption stalling?", "is this signal real or just noise?", "which of these causes explains the drop?", or before a headline conclusion ships. Not for factual single-answer questions (use `grounded-answer`) or when only one hypothesis exists.
- analyze-patent-claims
Dissects one patent's claim set the way an examiner reads it — independent versus dependent claims, the transitional phrase that fixes the scope (comprising = open, consisting of = closed), the elements of claim 1, the classification codes, and prior-art signals — and returns a structured record of what the filing actually covers. Use when a specific filing is on the table: "what does US00000000B2 cover?", "read this patent application", "freedom-to-operate on this claim", "is this patent-pending claim broad?". Not for a cluster of filings — use `read-patent-landscape` instead.
- analyze-release-notes
Turns a release note or changelog into a structured record — version and prior version, the SemVer bump, breaking changes and deprecations quoted verbatim, bug-fix count, CVE/GHSA advisories — and rates how much the release should worry a consumer. Use when a version delta is on the table: "what changed in v2.4?", "read this changelog", "any breaking changes before upgrading?", "what deprecations land in the next release?". Not for repository-level maintenance health — use `oss-project-health` instead.
- apply-hype-cycle
Places one discontinuous technology on the five phases of Gartner's Hype Cycle — Innovation Trigger, Peak of Inflated Expectations, Trough of Disillusionment, Slope of Enlightenment, Plateau of Productivity — by counting sourced indicators, then reports the phase, a time-to-plateau range and the stall risk. Use when asked "is X overhyped?", "what hype cycle stage is X at?", "has X come out of the trough of disillusionment?" or "how long until X is boring infrastructure?". Not for evidence of technical maturity (use `score-technology-readiness`) or one dated prediction (use `foresight`).
- assess-research-momentum
Reads a scholarly result set for one research area — publication S-curve phase, citation velocity and acceleration, author and institution concentration — and reports a research-front or maturing verdict with the observation that would flip it. Use when asked "is research on X heating up?", "is this a hot area or has it plateaued?", "what does the publication trend for X look like?" or "how much academic attention is X getting?". Not for market perception of a technology (use `apply-hype-cycle`) or for synthesising what the papers actually found (use `systematic-review`).
- assess-study-bias
Grades a randomised trial with Cochrane RoB 2 — the five domains of randomization process, deviations from intended interventions, missing outcome data, measurement of the outcome, and selection of the reported result — and emits per-domain and overall judgements of Low risk / Some concerns / High risk with the evidence cited. Use when a single study's trustworthiness is in question: "is this trial reliable?", "assess risk of bias in this RCT", "run RoB 2 on this study", "how much weight does this result deserve?". Not for grading a whole body of evidence (use `evidence-appraisal`) or for auditing an ML leaderboard claim (use `benchmark-model-claims`).
- backcasting
Works backwards from a desired end state to the present, producing a dated milestone chain, evidenced measure packages, a barrier and lock-in register and a near-term action agenda (Robinson's normative backcasting, Energy Policy, 1982). Use when the goal is a future that current trends will not deliver — "how does the city reach net-zero by 2040?", "backcast from the 2030 vision", "what must be true by {year} for {goal}?", "build a roadmap back from the target". Not for exploring which futures could happen (use `scenario-planning`) or for a probable-future forecast (use `foresight`).
- bayesian-update
Revises a stated probability in light of one new piece of evidence using the odds form of Bayes' rule — prior odds × Bayes factor = posterior odds — and reports the posterior in both relative and absolute terms with a sensitivity sweep over the prior and the likelihood ratio. Use when a belief has to move by a defensible amount: "update my prior", "how much should this evidence change my estimate?", "what's the posterior probability?", "run a Bayesian update on this". Not for enumerating and ranking several rival explanations (use `analysis-of-competing-hypotheses`) or for scoring forecasts after they resolve (use `brier-score-calibration`).
- benchmark-model-claims
Audits a claim about model or system performance against six integrity domains — test-set contamination, baseline pinning, seed and run variance, evaluator independence, metric selection, and whether the test set is large enough for the reported gap — and emits a 0–5 reliability score with named risk tags and a replication test. Use when a vendor post, paper or release note asserts a benchmark number: "91% on GSM8K", "outperforms GPT-4o", "SOTA on MMLU", "audit this benchmark claim". Not for grading a clinical or social-science trial (use `assess-study-bias`) and not for qualitative claims such as "better reasoning".
- brier-score-calibration
Scores a set of resolved probabilistic forecasts with the Brier score, decomposes it into reliability, resolution and uncertainty, compares it against the base-rate reference with a Brier Skill Score, and reports a per-bin calibration table showing where the forecaster is over- or underconfident. Use after predictions resolve — "how accurate were those forecasts?", "score my predictions", "was the panel overconfident?", "run a calibration check", "Brier score for this forecast set". Not for testing whether two groups differ (use `test-significance`) or for revising a single belief on new evidence (use `bayesian-update`).
- causal-layered-analysis
Deconstructs an issue through the four levels of Causal Layered Analysis — litany, systemic causes, worldview/discourse, myth/metaphor (Inayatullah, Futures, 1998) — then reconstructs a vertically coherent alternative future, producing a four-layer table, deconstruction notes and a reframed narrative. Use when framing itself is the obstacle: "why does this problem never get solved?", "reframe the debate on {issue}", "run a CLA on {issue}", "the scenarios all look like more of the same". Not for verifying a factual claim (use `grounded-fact-check`) or for numeric estimates (use `estimate-market-size`).
- cheapest-experiment
Turns a recommendation into its cheapest decisive test — the smallest bounded pilot, its cost and duration, a pass/fail rule fixed before the run, and what each outcome would change — in the spirit of Ries's Lean Startup build-measure-learn loop and MVP. Use when a brief ends in recommendations or a bet is about to be funded — "should this be piloted before committing?", "what is the cheapest experiment that would validate this bet?", "design an MVP test with a go/no-go rule", "how can this recommendation be de-risked?". Not for powered trials needing sample size — use `experimental-design`; not for enumerating how a chosen plan fails — use `premortem-analysis`.
- chemistry-claim-check
Checks molecular claims stated in prose — compound name against formula, degree of unsaturation, recomputed molecular weight, and the assessable part of the Lipinski rule of five — and returns a PASS/FAIL verdict per item with the arithmetic shown. Use when a paper, patent, or product page pairs a compound with a formula or a weight, on requests like "does this molecular weight check out?" or "is C2H6O2 really ethanol?". Not for validating a SMILES string — use `smiles-sanity-check` instead.
- cite-ieee
Converts a report's sources into IEEE numbered-bracket citations — inline `[1]`, `[1], [4]`, `[2]–[5]` markers numbered by first appearance, plus an end-of-document References list formatted per source type — and checks that the numbering and the list agree. Use when a document carries three or more sources and must look publishable: "format these references in IEEE style", "add citations to this report", "IEEE reference list", "renumber my citations". Not for checking whether a DOI or arXiv ID resolves (use `verify-citations`) and not for author-year styles such as APA.
- claim-provenance
Tags every fact-claim in a document inline as either `[validated, {source}]` or `[assumption, retire-by {milestone}]`, so a reader can see at a glance which sentences rest on evidence and which rest on reasoning, and what would convert each assumption into knowledge. Use when a brief mixes cited facts with reasoned-from-priors numbers — "mark which claims are sourced and which are assumptions", "tag the provenance of these claims", "which numbers here are actually validated?", "add validated/assumption tags to this draft". Not for verifying whether a number is correct (use `grounded-fact-check`) or for corroborating one claim across sources (use `triangulate-sources`).
- critique-report
Runs a structured self-critique of a finished draft before it is published — ten review points (does it answer the question, is the evidence sourced, is confidence honest, are limitations and counter-evidence stated, are the numbers defensible, is the next action actionable) plus three conditional ones — and returns a pass/fix/block verdict with located, specific fixes. Use when a draft is done but unsent: "review this brief before it goes out", "critique my report", "is this ready to publish?", "what would a reviewer catch?". Not for attacking one headline claim adversarially (use `red-team-claim`) or for checking reference identifiers (use `verify-citations`).
- cross-impact-analysis
Builds an N x N cross-impact matrix over a set of events or variables, rating each ordered pair's direct influence, then computes direct-stage MICMAC influence/dependence rankings that separate strategic levers from indicators and flags asymmetric pairs for re-rating (Gordon & Hayward, Futures, 1968; Godet's structural analysis). Use when asked "how do these trends affect each other?", "build a cross-impact matrix", "which of these drivers actually drive the system?", "run MICMAC on these variables" or "if X happens, does Y become more likely?". Not for eliciting the underlying estimates (use `delphi-method`) or writing scenario narratives (use `scenario-planning`).
- cynefin-classification
Classifies a decision situation into one of Cynefin's five contexts — Clear, Complicated, Complex, Chaotic or Disorder — and names the matching decision mode, so recommendations are framed as playbooks, expert analysis, safe-to-fail probes or stabilising action. Use before advising action under uncertainty — "classify this with Cynefin", "is there a best practice here or is this new territory?", "what kind of problem is this — known, knowable, or emergent?", "nobody agrees which kind of problem this even is". Not for enumerating a chosen plan's failure modes (use `premortem-analysis`) or designing the probe itself (use `cheapest-experiment`).
- decision-matrix-mcda
Ranks several options against weighted criteria with a decision matrix / multi-criteria decision analysis (Heuer & Pherson's Decision Matrix; Saaty's AHP pairwise comparisons for the weights, with a consistency check) and reports how robust the winner is to weight and score changes. Use when a choice has 3+ options and competing criteria — "which of these vendors/platforms/projects to pick?", "build a weighted decision matrix", "derive criteria weights with AHP pairwise comparisons", "score these options against the criteria". Not for ranking rival explanations of a cause (use `analysis-of-competing-hypotheses`) or picking the smallest validating test (use `cheapest-experiment`).
- decompose-research-question
Breaks a question too large, vague or multi-part to answer directly into a MECE issue tree of atomic sub-questions — each mapped to the skill or research step that answers it, ordered by dependency, with a named method for recombining the answers. Use when a request would otherwise sprawl — "help me understand the whole X landscape", "what's the future of Y and is it worth investing in?", "decompose this research question", "break this big question into parts before researching". Not for structuring an answer already in hand — use `pyramid-principle`; not for a single factual lookup — use `grounded-answer`.
- delphi-method
Aggregates independent expert judgments into a consensus estimate — or an honest map of where experts split — with the RAND Delphi method: anonymous panel, iterated rounds, controlled feedback of the group median and interquartile range, and a stability stopping rule. Use when data alone cannot answer and structured expert judgment is needed — "run a Delphi study", "expert consensus forecast on X", "when will X happen, according to the experts?", "expert elicitation of a date or probability". Not for a single reasoner's dated forecast — use `foresight`; not for branching futures — use `scenario-planning`; not for scoring forecasts afterwards — use `brier-score-calibration`.
- detect-funding-round
Parses a capital-raise announcement into a structured FundingEvent record — company, round stage, amount and currency, dates, lead and participating investors, valuation, use of proceeds, confidence. Use when a news item, press release or search result reports financing: "raised $47 million", "closed its Series B", "seed round", "post-money valuation of $400M", "led by Northgate Ventures". Not for change-of-control news — an acquisition or merger routes to `detect-ma-event`, a product launch to `analyze-release-notes`.
- detect-ma-event
Parses a news article, filing or press release about a change of control into a structured M&A event record — acquirer, target, deal type, consideration, deal value, premium, dates, regulatory jurisdictions, termination fee — and flags deal-structure risks. Use when text says "agreed to acquire", "merger of equals", "takeover bid", "all-cash deal", "go-private" or "divests its division". Not for primary capital raises — a seed or Series B round routes to `detect-funding-round`.
- estimate-market-size
Sizes a market by triangulating two independent estimates — a top-down cut from a larger market and a bottom-up build from buyer counts and revenue per buyer — and reports the TAM/SAM/SOM layer, a range, and whether the two estimates agree within an order of magnitude. Use when asked "how big is the X market?", "what's the TAM for this?", "is that multi-billion market number real?" or "size the opportunity before the board sees it". Not for recomputing one published figure's internal arithmetic (use `quantitative-sanity-check`) and not for industry attractiveness (use `five-forces-analysis`).
- estimative-language
Rewrites and lints the likelihood and confidence wording of an assessment against one declared standard — the ODNI ICD 203 estimative-probability terms, the UK PHIA Probability Yardstick, or the IPCC AR5 likelihood scale — converting numbers to calibrated terms, keeping confidence in its own sentence, and flagging vague words. Use when asked to "rewrite this assessment using ICD 203 probability language", "what does 'realistic possibility' mean in the UK yardstick", or "lint my brief for vague likelihood words and mixed confidence statements". Not for scoring how accurate past forecasts were, or for revising a probability after new information.
- evidence-appraisal
Rates how much certainty a body of evidence deserves, outcome by outcome, on the GRADE scale (Grading of Recommendations Assessment, Development and Evaluation — high, moderate, low, very low), naming the reason behind every downgrade or upgrade and writing the summary-of-findings row and a calibrated certainty statement. Use when asked "how strong is the evidence for X?", "GRADE this evidence", "is this proven or just one study?", or "the studies conflict — what is actually known?". Not for one study's risk of bias (use `assess-study-bias`) or for grading a non-study source such as an article or dataset (use `rate-source-admiralty`).
- evolution-stage
Tags a named technology, capability or vendor category with its Wardley evolution stage — Genesis, Custom built, Product (+rental) or Commodity (+utility) — anchored in observable evidence, and states the way of working that stage implies. Use when asked "is this a product or still bespoke?", "build vs buy for X", "where are these tools in their lifecycle?" or "what evolution stage is X at?". Not for the whole value chain and its dependencies (use `wardley-map-drafting`) and not for evidence of technical maturity (use `score-technology-readiness`).
- expected-value-decision-tree
Lays out a staged bet as a decision tree, rolls it back to expected values and prices the information a pilot or test would buy — expected value of perfect information (EVPI), expected value of sample information (EVSI) and a one-way sensitivity (tornado) with switching probabilities (Raiffa 1968). Use when a go/no-go, invest/wait or pilot-then-commit choice hinges on chance outcomes — "is the pilot worth running before committing?", "expected value of perfect information", "which uncertainty is the decision most sensitive to?", "build the decision tree". Not for designing the experiment itself (`cheapest-experiment`) or revising one probability (`bayesian-update`).
- experimental-design
Produces a preregisterable design card for a causal test before any data are collected — a falsifiable hypothesis with a minimum detectable effect, the randomization unit and mechanism, a power-based sample size per arm, a fixed stop rule and analysis plan, and an audit of the validity threats that survive randomization. Use when a test is being planned: "design an A/B test", "how many users per arm are needed?", "run a power analysis", "preregister this experiment", "is this study design sound?". Not for analysing a comparison whose data already exist (use `test-significance`) or for grading someone else's finished trial (use `assess-study-bias`).
- five-forces-analysis
Assesses the structural profitability of an industry with Porter's Five Forces — entry threat, supplier power, buyer power, substitutes and rivalry — rating each on evidence, naming the one or two forces that actually control profits, and reporting how they are trending. Use when the question is whether a market is worth entering or staying in — "how attractive is this industry?", "Porter five forces for this market", "what are the barriers to entry here?", "why are margins in this sector so thin?". Not for placing one firm against its rivals (use `position-competitor`) or for sizing demand (use `estimate-market-size`).
- foresight
Produces one dated, falsifiable prediction about a single technology, trend or market shift — the milestone with a date and a confidence, its accelerants and blockers with lead times, weak signals to watch now, kill-signals that would retract it, and a review date. Use for timing questions — "when will X happen?", "is this accelerating or stalling?", "what should be watched to know if this is real?", "give a dated forecast with kill-signals". Not for branching what-could-happen questions — use `scenario-planning`; not for a standing scan of many signals — use `horizon-scanning`; not for polling an expert panel — use `delphi-method`.
- futures-wheel
Maps the consequences of one defined change ring by ring — primary, secondary and tertiary impacts, each chained to a named parent with its mechanism — and returns an assessed consequence map with convergence points, a desirable/undesirable valence check and an implication-to-action table (Jerome C. Glenn's Futures Wheel, 1971). Use when asked "what are the second-order effects of {change}?", "map the implications of {decision}", "futures wheel", "unintended consequences" or "and then what happens?". Not for how likely a consequence is (use `bayesian-update`) or for branching alternative futures (use `scenario-planning`).
- grounded-answer
Produces a factual answer in which every claim has been independently checked against a source and cited, using Chain-of-Verification (CoVe — Dhuliawala et al., 2023) — the answer ships only with claims that survived verification, plus a list of what could not be verified. Use when a factual answer will reach a reader or a decision — "give me a sourced answer", "what year did that happen? verify it", "run chain-of-verification on this question", "answer with citations, no guessing". Not for auditing the numbers in an already-written draft (use `grounded-fact-check`) or for interpretive questions where several explanations compete (use `analysis-of-competing-hypotheses`).
- grounded-fact-check
Audits the load-bearing specifics of a finished draft before it is published — the qubit counts, market figures, benchmark scores, dates and precedence claims a reader would act on — verifying each against live sources with a neutral query and returning a ledger of confirmed, corrected, unverifiable claims. Use when a draft is written and its numbers must hold up — "fact-check this report before it ships", "are these figures right?", "check the numbers in this brief", "verify the specifics in this draft". Not for producing a fresh sourced answer (use `grounded-answer`) or recomputing a document's internal arithmetic (use `quantitative-sanity-check`).
- high-impact-low-probability
Takes one event judged unlikely but devastating, postulates it has happened, and reasons backwards to the pathways that could produce it — returning the impact, 2–4 distinct pathways with their triggers, observable early-warning indicators, and a proportionate hedge (High-Impact/Low-Probability and "What If?" Analysis, US Tradecraft Primer 2009). Use when a severe scenario is being dismissed as improbable — "what if the sole supplier were cut off?", "stress-test how this unlikely event could actually happen", "run a high-impact/low-probability analysis". Not for a decision assumed to have failed (`premortem-analysis`) or a set of plausible futures (`scenario-planning`).
- horizon-scanning
Runs a horizon scan — a systematic early-warning sweep of sources outside an organisation's usual reading for weak signals and emerging issues — and delivers a prioritised horizon-scanning report with impact/likelihood ratings, verified sources and next steps (UK Government Office for Science Futures Toolkit; Amanatidou et al.). Use when asked "what's on the horizon for {domain} over the next ten years?", "what should the strategy team be watching?", "scan for weak signals", "what could blindside us?" or "set up horizon scanning". Not for a dated prediction on one named trend (use `foresight`) or a rated inventory of macro-environment factors (use `steep-pestle-analysis`).
- indicators-validation
Builds and validates a signpost list — indicators of change — for rival scenarios or hypotheses (Heuer & Pherson's Indicators and Indicators Validator), rating every indicator against every scenario, dropping those that would appear under all of them, and returning the diagnostic set with a monitoring plan. Use when the scenarios already exist and the question is which is unfolding — "what early-warning indicators would tell us which scenario is unfolding?", "validate the indicator list", "build a signpost list for the scenario set", "are these indicators actually diagnostic?". Not for building the scenarios (use `scenario-planning`) or an open-ended environmental scan (`horizon-scanning`).
+30
Using structured-analytic-skills in Claude Code
Skills need no command. Claude reads them when your request matches what they describe.
Update, turn off or remove
- Update the plugin (shell)
- claude plugin update structured-analytic-skills@structured-analytic-skills
- Refresh the marketplace
- /plugin marketplace update structured-analytic-skills
- Turn it off for now
- /plugin disable structured-analytic-skills@structured-analytic-skills
- Remove it
- /plugin uninstall structured-analytic-skills@structured-analytic-skills
Skills in structured-analytic-skills
- abstain-or-escalateradarist/structured-analytic-skills
Decides what to do with a claim that verification could not confirm — drop it, refuse it in plain words, escalate it as a named open question, or report the disagreement between conflicting sources — instead of shipping a hedged guess. Use when verification has already failed: "there is no source for this", "the sources contradict each other", "should this claim be dropped or flagged?", "abstain rather than guess". Not for running the verification pass itself (use `grounded-answer`) or for checking published numbers against live sources (use `grounded-fact-check`).
3 files8stars - amstar2-review-appraisalradarist/structured-analytic-skills
Appraises a systematic review or meta-analysis with AMSTAR 2 (AMSTAR-2; Shea et al., BMJ 2017): answers the 16 checklist items from the review's report, flags flaws in the seven critical domains and derives the overall confidence rating — High, Moderate, Low or Critically low — never a summed score. Use when deciding whether a published review can be relied on, or ranking reviews for an umbrella review: "appraise this systematic review with AMSTAR 2", "how trustworthy is this meta-analysis?", "which AMSTAR 2 items are critical?". Not for one trial (use `assess-study-bias`), a body of evidence per outcome (use `evidence-appraisal`) or conducting a review (use `systematic-review`).
5 files8stars - analysis-of-competing-hypothesesradarist/structured-analytic-skills
Ranks rival explanations of a question by which one the evidence contradicts least (Heuer's Analysis of Competing Hypotheses — ACH — CIA), then names the single piece of evidence the ranking hinges on. Use when a question has two or more plausible explanations and confirmation bias would be costly — "why is adoption stalling?", "is this signal real or just noise?", "which of these causes explains the drop?", or before a headline conclusion ships. Not for factual single-answer questions (use `grounded-answer`) or when only one hypothesis exists.
5 files8stars - analyze-patent-claimsradarist/structured-analytic-skills
Dissects one patent's claim set the way an examiner reads it — independent versus dependent claims, the transitional phrase that fixes the scope (comprising = open, consisting of = closed), the elements of claim 1, the classification codes, and prior-art signals — and returns a structured record of what the filing actually covers. Use when a specific filing is on the table: "what does US00000000B2 cover?", "read this patent application", "freedom-to-operate on this claim", "is this patent-pending claim broad?". Not for a cluster of filings — use `read-patent-landscape` instead.
5 files8stars - analyze-release-notesradarist/structured-analytic-skills
Turns a release note or changelog into a structured record — version and prior version, the SemVer bump, breaking changes and deprecations quoted verbatim, bug-fix count, CVE/GHSA advisories — and rates how much the release should worry a consumer. Use when a version delta is on the table: "what changed in v2.4?", "read this changelog", "any breaking changes before upgrading?", "what deprecations land in the next release?". Not for repository-level maintenance health — use `oss-project-health` instead.
5 files8stars - apply-hype-cycleradarist/structured-analytic-skills
Places one discontinuous technology on the five phases of Gartner's Hype Cycle — Innovation Trigger, Peak of Inflated Expectations, Trough of Disillusionment, Slope of Enlightenment, Plateau of Productivity — by counting sourced indicators, then reports the phase, a time-to-plateau range and the stall risk. Use when asked "is X overhyped?", "what hype cycle stage is X at?", "has X come out of the trough of disillusionment?" or "how long until X is boring infrastructure?". Not for evidence of technical maturity (use `score-technology-readiness`) or one dated prediction (use `foresight`).
4 files8stars - assess-research-momentumradarist/structured-analytic-skills
Reads a scholarly result set for one research area — publication S-curve phase, citation velocity and acceleration, author and institution concentration — and reports a research-front or maturing verdict with the observation that would flip it. Use when asked "is research on X heating up?", "is this a hot area or has it plateaued?", "what does the publication trend for X look like?" or "how much academic attention is X getting?". Not for market perception of a technology (use `apply-hype-cycle`) or for synthesising what the papers actually found (use `systematic-review`).
6 files8stars - assess-study-biasradarist/structured-analytic-skills
Grades a randomised trial with Cochrane RoB 2 — the five domains of randomization process, deviations from intended interventions, missing outcome data, measurement of the outcome, and selection of the reported result — and emits per-domain and overall judgements of Low risk / Some concerns / High risk with the evidence cited. Use when a single study's trustworthiness is in question: "is this trial reliable?", "assess risk of bias in this RCT", "run RoB 2 on this study", "how much weight does this result deserve?". Not for grading a whole body of evidence (use `evidence-appraisal`) or for auditing an ML leaderboard claim (use `benchmark-model-claims`).
5 files8stars - backcastingradarist/structured-analytic-skills
Works backwards from a desired end state to the present, producing a dated milestone chain, evidenced measure packages, a barrier and lock-in register and a near-term action agenda (Robinson's normative backcasting, Energy Policy, 1982). Use when the goal is a future that current trends will not deliver — "how does the city reach net-zero by 2040?", "backcast from the 2030 vision", "what must be true by {year} for {goal}?", "build a roadmap back from the target". Not for exploring which futures could happen (use `scenario-planning`) or for a probable-future forecast (use `foresight`).
4 files8stars - bayesian-updateradarist/structured-analytic-skills
Revises a stated probability in light of one new piece of evidence using the odds form of Bayes' rule — prior odds × Bayes factor = posterior odds — and reports the posterior in both relative and absolute terms with a sensitivity sweep over the prior and the likelihood ratio. Use when a belief has to move by a defensible amount: "update my prior", "how much should this evidence change my estimate?", "what's the posterior probability?", "run a Bayesian update on this". Not for enumerating and ranking several rival explanations (use `analysis-of-competing-hypotheses`) or for scoring forecasts after they resolve (use `brier-score-calibration`).
5 files8stars - benchmark-model-claimsradarist/structured-analytic-skills
Audits a claim about model or system performance against six integrity domains — test-set contamination, baseline pinning, seed and run variance, evaluator independence, metric selection, and whether the test set is large enough for the reported gap — and emits a 0–5 reliability score with named risk tags and a replication test. Use when a vendor post, paper or release note asserts a benchmark number: "91% on GSM8K", "outperforms GPT-4o", "SOTA on MMLU", "audit this benchmark claim". Not for grading a clinical or social-science trial (use `assess-study-bias`) and not for qualitative claims such as "better reasoning".
4 files8stars - brier-score-calibrationradarist/structured-analytic-skills
Scores a set of resolved probabilistic forecasts with the Brier score, decomposes it into reliability, resolution and uncertainty, compares it against the base-rate reference with a Brier Skill Score, and reports a per-bin calibration table showing where the forecaster is over- or underconfident. Use after predictions resolve — "how accurate were those forecasts?", "score my predictions", "was the panel overconfident?", "run a calibration check", "Brier score for this forecast set". Not for testing whether two groups differ (use `test-significance`) or for revising a single belief on new evidence (use `bayesian-update`).
6 files8stars - causal-layered-analysisradarist/structured-analytic-skills
Deconstructs an issue through the four levels of Causal Layered Analysis — litany, systemic causes, worldview/discourse, myth/metaphor (Inayatullah, Futures, 1998) — then reconstructs a vertically coherent alternative future, producing a four-layer table, deconstruction notes and a reframed narrative. Use when framing itself is the obstacle: "why does this problem never get solved?", "reframe the debate on {issue}", "run a CLA on {issue}", "the scenarios all look like more of the same". Not for verifying a factual claim (use `grounded-fact-check`) or for numeric estimates (use `estimate-market-size`).
3 files8stars - cheapest-experimentradarist/structured-analytic-skills
Turns a recommendation into its cheapest decisive test — the smallest bounded pilot, its cost and duration, a pass/fail rule fixed before the run, and what each outcome would change — in the spirit of Ries's Lean Startup build-measure-learn loop and MVP. Use when a brief ends in recommendations or a bet is about to be funded — "should this be piloted before committing?", "what is the cheapest experiment that would validate this bet?", "design an MVP test with a go/no-go rule", "how can this recommendation be de-risked?". Not for powered trials needing sample size — use `experimental-design`; not for enumerating how a chosen plan fails — use `premortem-analysis`.
3 files8stars - chemistry-claim-checkradarist/structured-analytic-skills
Checks molecular claims stated in prose — compound name against formula, degree of unsaturation, recomputed molecular weight, and the assessable part of the Lipinski rule of five — and returns a PASS/FAIL verdict per item with the arithmetic shown. Use when a paper, patent, or product page pairs a compound with a formula or a weight, on requests like "does this molecular weight check out?" or "is C2H6O2 really ethanol?". Not for validating a SMILES string — use `smiles-sanity-check` instead.
4 files8stars - cite-ieeeradarist/structured-analytic-skills
Converts a report's sources into IEEE numbered-bracket citations — inline `[1]`, `[1], [4]`, `[2]–[5]` markers numbered by first appearance, plus an end-of-document References list formatted per source type — and checks that the numbering and the list agree. Use when a document carries three or more sources and must look publishable: "format these references in IEEE style", "add citations to this report", "IEEE reference list", "renumber my citations". Not for checking whether a DOI or arXiv ID resolves (use `verify-citations`) and not for author-year styles such as APA.
7 files8stars - claim-provenanceradarist/structured-analytic-skills
Tags every fact-claim in a document inline as either `[validated, {source}]` or `[assumption, retire-by {milestone}]`, so a reader can see at a glance which sentences rest on evidence and which rest on reasoning, and what would convert each assumption into knowledge. Use when a brief mixes cited facts with reasoned-from-priors numbers — "mark which claims are sourced and which are assumptions", "tag the provenance of these claims", "which numbers here are actually validated?", "add validated/assumption tags to this draft". Not for verifying whether a number is correct (use `grounded-fact-check`) or for corroborating one claim across sources (use `triangulate-sources`).
3 files8stars - critique-reportradarist/structured-analytic-skills
Runs a structured self-critique of a finished draft before it is published — ten review points (does it answer the question, is the evidence sourced, is confidence honest, are limitations and counter-evidence stated, are the numbers defensible, is the next action actionable) plus three conditional ones — and returns a pass/fix/block verdict with located, specific fixes. Use when a draft is done but unsent: "review this brief before it goes out", "critique my report", "is this ready to publish?", "what would a reviewer catch?". Not for attacking one headline claim adversarially (use `red-team-claim`) or for checking reference identifiers (use `verify-citations`).
4 files8stars - cross-impact-analysisradarist/structured-analytic-skills
Builds an N x N cross-impact matrix over a set of events or variables, rating each ordered pair's direct influence, then computes direct-stage MICMAC influence/dependence rankings that separate strategic levers from indicators and flags asymmetric pairs for re-rating (Gordon & Hayward, Futures, 1968; Godet's structural analysis). Use when asked "how do these trends affect each other?", "build a cross-impact matrix", "which of these drivers actually drive the system?", "run MICMAC on these variables" or "if X happens, does Y become more likely?". Not for eliciting the underlying estimates (use `delphi-method`) or writing scenario narratives (use `scenario-planning`).
5 files8stars - cynefin-classificationradarist/structured-analytic-skills
Classifies a decision situation into one of Cynefin's five contexts — Clear, Complicated, Complex, Chaotic or Disorder — and names the matching decision mode, so recommendations are framed as playbooks, expert analysis, safe-to-fail probes or stabilising action. Use before advising action under uncertainty — "classify this with Cynefin", "is there a best practice here or is this new territory?", "what kind of problem is this — known, knowable, or emergent?", "nobody agrees which kind of problem this even is". Not for enumerating a chosen plan's failure modes (use `premortem-analysis`) or designing the probe itself (use `cheapest-experiment`).
3 files8stars - decision-matrix-mcdaradarist/structured-analytic-skills
Ranks several options against weighted criteria with a decision matrix / multi-criteria decision analysis (Heuer & Pherson's Decision Matrix; Saaty's AHP pairwise comparisons for the weights, with a consistency check) and reports how robust the winner is to weight and score changes. Use when a choice has 3+ options and competing criteria — "which of these vendors/platforms/projects to pick?", "build a weighted decision matrix", "derive criteria weights with AHP pairwise comparisons", "score these options against the criteria". Not for ranking rival explanations of a cause (use `analysis-of-competing-hypotheses`) or picking the smallest validating test (use `cheapest-experiment`).
5 files8stars - decompose-research-questionradarist/structured-analytic-skills
Breaks a question too large, vague or multi-part to answer directly into a MECE issue tree of atomic sub-questions — each mapped to the skill or research step that answers it, ordered by dependency, with a named method for recombining the answers. Use when a request would otherwise sprawl — "help me understand the whole X landscape", "what's the future of Y and is it worth investing in?", "decompose this research question", "break this big question into parts before researching". Not for structuring an answer already in hand — use `pyramid-principle`; not for a single factual lookup — use `grounded-answer`.
3 files8stars - delphi-methodradarist/structured-analytic-skills
Aggregates independent expert judgments into a consensus estimate — or an honest map of where experts split — with the RAND Delphi method: anonymous panel, iterated rounds, controlled feedback of the group median and interquartile range, and a stability stopping rule. Use when data alone cannot answer and structured expert judgment is needed — "run a Delphi study", "expert consensus forecast on X", "when will X happen, according to the experts?", "expert elicitation of a date or probability". Not for a single reasoner's dated forecast — use `foresight`; not for branching futures — use `scenario-planning`; not for scoring forecasts afterwards — use `brier-score-calibration`.
8 files8stars - detect-funding-roundradarist/structured-analytic-skills
Parses a capital-raise announcement into a structured FundingEvent record — company, round stage, amount and currency, dates, lead and participating investors, valuation, use of proceeds, confidence. Use when a news item, press release or search result reports financing: "raised $47 million", "closed its Series B", "seed round", "post-money valuation of $400M", "led by Northgate Ventures". Not for change-of-control news — an acquisition or merger routes to `detect-ma-event`, a product launch to `analyze-release-notes`.
4 files8stars - detect-ma-eventradarist/structured-analytic-skills
Parses a news article, filing or press release about a change of control into a structured M&A event record — acquirer, target, deal type, consideration, deal value, premium, dates, regulatory jurisdictions, termination fee — and flags deal-structure risks. Use when text says "agreed to acquire", "merger of equals", "takeover bid", "all-cash deal", "go-private" or "divests its division". Not for primary capital raises — a seed or Series B round routes to `detect-funding-round`.
4 files8stars - estimate-market-sizeradarist/structured-analytic-skills
Sizes a market by triangulating two independent estimates — a top-down cut from a larger market and a bottom-up build from buyer counts and revenue per buyer — and reports the TAM/SAM/SOM layer, a range, and whether the two estimates agree within an order of magnitude. Use when asked "how big is the X market?", "what's the TAM for this?", "is that multi-billion market number real?" or "size the opportunity before the board sees it". Not for recomputing one published figure's internal arithmetic (use `quantitative-sanity-check`) and not for industry attractiveness (use `five-forces-analysis`).
4 files8stars - estimative-languageradarist/structured-analytic-skills
Rewrites and lints the likelihood and confidence wording of an assessment against one declared standard — the ODNI ICD 203 estimative-probability terms, the UK PHIA Probability Yardstick, or the IPCC AR5 likelihood scale — converting numbers to calibrated terms, keeping confidence in its own sentence, and flagging vague words. Use when asked to "rewrite this assessment using ICD 203 probability language", "what does 'realistic possibility' mean in the UK yardstick", or "lint my brief for vague likelihood words and mixed confidence statements". Not for scoring how accurate past forecasts were, or for revising a probability after new information.
5 files8stars - evidence-appraisalradarist/structured-analytic-skills
Rates how much certainty a body of evidence deserves, outcome by outcome, on the GRADE scale (Grading of Recommendations Assessment, Development and Evaluation — high, moderate, low, very low), naming the reason behind every downgrade or upgrade and writing the summary-of-findings row and a calibrated certainty statement. Use when asked "how strong is the evidence for X?", "GRADE this evidence", "is this proven or just one study?", or "the studies conflict — what is actually known?". Not for one study's risk of bias (use `assess-study-bias`) or for grading a non-study source such as an article or dataset (use `rate-source-admiralty`).
4 files8stars - evolution-stageradarist/structured-analytic-skills
Tags a named technology, capability or vendor category with its Wardley evolution stage — Genesis, Custom built, Product (+rental) or Commodity (+utility) — anchored in observable evidence, and states the way of working that stage implies. Use when asked "is this a product or still bespoke?", "build vs buy for X", "where are these tools in their lifecycle?" or "what evolution stage is X at?". Not for the whole value chain and its dependencies (use `wardley-map-drafting`) and not for evidence of technical maturity (use `score-technology-readiness`).
4 files8stars - expected-value-decision-treeradarist/structured-analytic-skills
Lays out a staged bet as a decision tree, rolls it back to expected values and prices the information a pilot or test would buy — expected value of perfect information (EVPI), expected value of sample information (EVSI) and a one-way sensitivity (tornado) with switching probabilities (Raiffa 1968). Use when a go/no-go, invest/wait or pilot-then-commit choice hinges on chance outcomes — "is the pilot worth running before committing?", "expected value of perfect information", "which uncertainty is the decision most sensitive to?", "build the decision tree". Not for designing the experiment itself (`cheapest-experiment`) or revising one probability (`bayesian-update`).
6 files8stars
structured-analytic-skills plugin questions
How do I install the structured-analytic-skills plugin?
In Claude Code run /plugin marketplace add radarist/structured-analytic-skills, then /plugin install structured-analytic-skills@structured-analytic-skills.
What does the structured-analytic-skills plugin include?
structured-analytic-skills bundles 70 skills.
How do I update or remove the structured-analytic-skills plugin?
Update it with claude plugin update structured-analytic-skills@structured-analytic-skills in your shell, or from the Installed tab of /plugin. Remove it with /plugin uninstall structured-analytic-skills@structured-analytic-skills.
What license is the structured-analytic-skills plugin under?
MIT, as stated in the plugin's manifest.