Skip to content

rampstackco/claude-skills / skills/evidence-based-reviews

evidence-based-reviews

Verified

Produce honest product reviews and buying guides without fabricated first-hand experience, using disclosed evidence tiers: verified specs, owner-experience synthesis at scale, expert triangulation, and hands-on only when true. Use this skill whenever the user wants to write product reviews or buying guides without hands-on access, set up a review methodology, add evidence disclosure to review content, align reviews with Google's reviews system or FTC affiliate disclosure expectations, or decide when Review and Product schema are honest to use. Triggers on product review, buying guide, best-of list, review methodology, evidence basis, hands-on testing, we tested, affiliate review site, review disclosure, methodology block, original research, reviews system. Also triggers when review content claims testing that did not happen, or when an affiliate site needs a trust mechanism that survives scrutiny.

FileKindSize
README.mdmarkdown363 B
SKILL.mdskill md12.1 KB
references/evidence-tiers.mdmarkdown4.1 KB
references/methodology-block-template.mdmarkdown2.7 KB

references/evidence-tiers.md

View on GitHub
# The four evidence tiers, in depth

The tiers are not a ladder of prestige; they are different kinds of knowledge with different verification work and different claims they can support. A strong piece usually combines two or three, and says which.

## Tier 1: verified manufacturer specs

**The verification step.** Cite the specification to the manufacturer's own current product page. Aggregators, retailer listings, and other reviews drift: units get converted wrong, superseded model years linger, and one site's typo becomes the category's consensus number. When the manufacturer's page and a retailer listing disagree, note the discrepancy in the piece; the disagreement is information.

**Date the pull.** Specs change across model years. A spec table that says when its numbers were verified survives the next revision; one that does not becomes silently wrong.

**What tier 1 supports:** factual comparison (weights, dimensions, capacities, materials), compatibility and fit answers, price-tier groupings. **What it cannot support:** any quality verdict. A spec sheet cannot tell you the thing rattles.

## Tier 2: owner-experience synthesis at scale

The workhorse tier for review work without hands-on access, and the one most often done dishonestly. The honest version:

**Scale and candor.** Read across sources: retailer review corpora, owner forums, warranty and return discussions. State the scale in the piece ("across roughly 1,400 retailer reviews and two seasons of forum threads"). If the sample is small, say so and weaken the claim to match.

**Recurrence over volume.** The finding is the pattern that recurs across independent sources, not the most dramatic single report. One catastrophic failure post is an anecdote; the same failure described by unrelated owners across two sites is a finding.

**Splits are findings.** When owners divide (half love the trigger, half send it back), the split is the most useful thing the piece can report, usually with the variable that explains it (hand size, draw weight, use case).

**No cherry-picking.** Deciding the verdict first and quoting only supporting owners is fabrication with extra steps. The synthesis is written from the pattern, not toward a pick.

**Name the corpora.** Which retailers, which communities, what period. This is what makes the synthesis auditable and what makes it original analysis rather than vibes.

## Tier 3: expert-source triangulation

**Named sources only.** "Experts agree" without names is decoration. Name the publication or tester, and link where possible.

**Triangulate, never average.** Three credible sources agreeing is strong signal; two credible sources disagreeing is stronger content. Surface the disagreement and look for the condition that explains it: different test protocols, different units, different use assumptions. Averaging opposite verdicts into "mixed reviews" throws away exactly what the reader came for.

**Independence check.** Three reviews that all trace to the same press kit are one source wearing three hats.

## Tier 4: hands-on, only when true

**Quantify the experience.** "We shot it" hides more than it tells. The honest form states extent and conditions: how many arrows or hours or miles, over what period, in what setting, with what comparison points. The quantification is what separates a real claim from a vibe.

**Flag it.** On a site where synthesis is the norm, hands-on time is the exception worth marking. Flagging it does double duty: it strengthens the piece that has it and keeps every other piece honest by contrast.

**The upgrade-and-mark pattern.** When hands-on experience arrives for a piece originally published on tiers 1 to 3:

1. Update the piece with the hands-on findings.
2. Update the methodology block: the hands-on line changes from "none claimed" to the quantified statement.
3. Mark the update with the date and what changed ("Updated [date]: added hands-on findings after six weeks with the production unit; the durability verdict held, the comfort note changed").
4. Never rewrite history. The piece does not pretend it was always hands-on; the visible upgrade is itself evidence the site's claims mean something.
finishing-a-development-branchobra/superpowersUse when implementation is complete, all tests pass, and you need to decide how to integrate the workStars: 322kUpdated Claude CodeCodexVerifiedInstall
systematic-debuggingobra/superpowersUse when encountering any bug, test failure, or unexpected behavior, before proposing fixesStars: 322kUpdated Claude CodeCodexVerifiedInstall
test-driven-developmentobra/superpowersUse when implementing any feature or bugfix, before writing implementation codeStars: 322kUpdated Claude CodeCodexVerifiedInstall
skill-creatoranthropics/skillsCreate new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy.Stars: 198kUpdated Claude CodeCodexVerifiedInstall