Stats
Actions
Tags
From strategy-consultant
Pressure-test research findings and analytical conclusions for robustness. Use when someone asks to "sense-check this", "pressure-test these findings", "challenge this analysis", "stress-test the argument", "play devil's advocate", "poke holes in this", or when the analytical workflow needs a quality gate between research and synthesis. Also trigger when someone presents a conclusion and wants it challenged before taking it to a client.
How this skill is triggered — by the user, by Claude, or both
Slash command
/strategy-consultant:sense-checkThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Take the research findings and emerging conclusions and systematically try to break them. This is the quality gate between having evidence and having a defensible argument. Most weak consulting work fails here — the analyst found data that supports the hypothesis and stopped looking.
Take the research findings and emerging conclusions and systematically try to break them. This is the quality gate between having evidence and having a defensible argument. Most weak consulting work fails here — the analyst found data that supports the hypothesis and stopped looking.
List every substantive claim that the eventual storyline will rest on. For each claim, note:
For each major claim, check whether the same conclusion holds when approached from different angles:
If three independent angles point the same way, the claim is on solid ground. If only one does, flag it as fragile.
This step has two parts. Both are mandatory.
Part A: Steel-man the analytical counter-argument. For each key conclusion, construct the strongest possible counter-argument:
Write the steel-man argument out in full. If you cannot articulate a strong opposing case, the conclusion may be trivially true (and therefore not very useful) — or you have not thought hard enough.
Part B: Systematically extract client-raised objections. Re-read the client call transcript, brief, and all source materials. Extract every concern, hesitation, objection, or risk the client raised — even offhand comments. For each one:
Client-raised concerns that go unaddressed in the deliverable are a credibility risk. The client will notice their own question was not answered. Even if the concern seems minor, it is safer to address it briefly than to ignore it.
Part C: Evidence-implied risks. Review the research findings for risks, threats, or caveats that the client did not raise but that the evidence itself surfaces. For each:
For every quantitative claim:
This is the most important check in the sense-check process. Re-read the Precision Anchor from the problem-definition phase, then ask:
Has the analysis drifted? Compare the emerging conclusion to the exact question in the Precision Anchor. Common drift patterns:
Is the answer at the right altitude? This is the most common and most dangerous form of analytical drift. The analysis may address the right topic but at the wrong level of specificity:
An altitude mismatch is NOT a precise answer. It is a PARTIAL answer at best. When the sense-check identifies an altitude mismatch, the verdict must be PARTIAL and the gap must be explicitly stated: "The evidence answers [what it actually answers] but the client asked for [what they actually asked]. The gap is [specific missing data]. To close it, [specific recommended action]."
Do NOT let altitude mismatches pass as precise answers. A consultant who presents network totals when the client asked for per-store breakdowns has not answered the question — they have provided useful context that happens to be on the same topic.
If the data doesn't match the question: This is NOT a reason to answer a different question. Instead, the sense-check must flag: (a) what the available data can answer, (b) the gap between that and what the client asked, and (c) a clear recommendation for how to close the gap (additional research, client data, expert interviews). If the gap cannot be closed, the conclusion must include an explicit qualification.
Coverage completeness check: If the Precision Anchor includes a Deliverable Blueprint with Coverage Dimensions, produce a Coverage Matrix that maps every dimension the client asked for against the evidence base. The matrix should make gaps immediately visible. Any dimension classified as GAP downgrades the overall verdict to PARTIAL at best — you cannot call an answer PRECISE if entire dimensions the client asked for are missing.
Verdict: [The emerging conclusion PRECISELY answers the Precision Anchor question / The conclusion PARTIALLY answers it — specify the gap / The analysis has DRIFTED — specify to what question]
If the verdict is DRIFTED, recommend looping back to targeted research before proceeding to synthesis. Do not let a drifted analysis pass this gate.
The preceding steps ensure the analysis precisely answers the question. This step checks whether the deliverable stops at precision when it should go further.
Review the full research base and ask: "Is there anything the research uncovered that a knowledgeable practitioner would expect to see in a deliverable on this topic — even though the client didn't ask for it?"
This is not about padding the report. Clients often don't know what to ask for, and part of a consultant's value is surfacing what they need to know beyond what they asked.
Scan the validated findings, expert extractions, and any [ADJACENT] findings. For each, ask: would omitting this leave the client materially less equipped to act? If yes, flag it for inclusion as a value-add section in synthesis.
Run through this checklist:
Confirmation bias: Did the research primarily seek evidence that supports the hypothesis? Is there a systematic gap in disconfirming evidence?
Survivorship bias: Are we only looking at successful examples? What about companies/markets/strategies that tried the same thing and failed?
Cherry-picking: Are there data points that were found but excluded because they did not fit the narrative?
False precision: Are we presenting uncertain estimates as if they were exact?
Missing "compared to what?": Every number needs context. 8% growth sounds good — unless the market grew 12%. $50M in savings sounds large — unless the cost base is $20B.
Recency bias: Are we over-weighting recent events and under-weighting structural factors?
Anchoring: Are we anchored on the first number we found rather than the best number?
## Overall Assessment
[1-2 sentences: How robust is the evidence base? Can we proceed to synthesis or does more work need to happen?]
## Question-Answer Precision Verdict: [PRECISE / PARTIAL / DRIFTED]
[If PARTIAL: what does the data answer vs. what was asked? What is the gap?]
[If DRIFTED: what question is the analysis actually answering? What would bring it back on track?]
## Claim-by-Claim Assessment
| Claim | Evidence Strength | Triangulated? | Counter-Argument | Verdict |
|-------|------------------|--------------|-----------------|---------|
## Coverage Matrix (if Deliverable Blueprint has Coverage Dimensions)
| Dimension Unit | Status (COVERED / THIN / GAP) | Notes |
|----------------|-------------------------------|-------|
## Key Vulnerabilities
[The 2-3 weakest points in the argument — where a skeptical partner or client would push back]
## Steel-Man Counter-Narrative
[The best case against the emerging conclusion, written as a coherent argument]
## Client-Raised Objections Inventory
| # | Client Concern | Source (where raised) | Addressed in Evidence? | Action Needed |
|---|---------------|----------------------|----------------------|---------------|
[Every concern, hesitation, or objection the client raised in the call, brief, or messages. For each: note whether the current evidence addresses it, and if not, what action is needed to address it in the synthesis/report.]
## Evidence-Implied Risks
| # | Risk | Evidence Source | Addressed in Storyline? | Action Needed |
|---|------|----------------|------------------------|---------------|
[Risks, threats, or caveats that the client did not raise but that the research findings themselves surface. For each: the risk, the evidence that implies it, whether the current storyline addresses it, and if not, what action is needed.]
## Value-Add Opportunities
[High-value findings from the research base that go beyond the Precision Anchor but would materially elevate the deliverable. For each: the finding, why it matters to the client, and whether the current storyline would include or filter it out.]
## Math Checks
[Results of back-of-envelope sanity checks on key numbers]
## Analytical Traps Identified
[Any biases or logical errors found, with specific examples]
## Recommendation
[Proceed to synthesis / Revisit specific research areas / Downgrade confidence on specific claims / REDIRECT — analysis has drifted, needs targeted research on the actual question]
Present the sense-check report to the user. If vulnerabilities are serious or the precision verdict is DRIFTED, recommend specific follow-up research before proceeding to synthesis. A DRIFTED verdict is a hard stop — do not proceed to synthesis until the analysis is redirected to answer the actual question.
npx claudepluginhub chipalexandru/strategy-consultantQA data analyses for methodology, accuracy, biases, and pitfalls before stakeholder sharing. Spot-checks calculations, SQL results, visualizations, and conclusions.
Analyzes claims by mapping arguments, auditing evidence quality, detecting logical fallacies and biases, and issuing verdicts. For evaluating research or technical arguments.
Detects statistical errors, logical fallacies, and methodological issues in research content. Use for validating statistics, auditing quantitative claims, or checking methodology.