Skills

104 skills · 5 measured · 4 tested and cut

104 Claude skills. 5 carry a number.

Search them, take them, no account. Every card says whether we measured it or not. How we test →

All 104 shown

5 skills with a measured result

Run through the harness against a control. The numbers, the p value and the cost of using it are on each page.

Code reviewMeasured Blast radius review Treats the bug in front of you as one member of a family, then goes and finds the rest of the family. The search discipline is the asset. skill · 2,842 words · MIT Read the result

MarketingMeasured Email deliverability audit Reads SPF, DKIM and DMARC as one system rather than three records, then works down through list quality, sending pattern and message construction to a ship or hold. skill · 4,826 words · MIT Read the result

Frontend designMeasured Form design and validation audit Checks every field for the right autocomplete token, the right keyboard, validation that fires on blur rather than on the first keystroke, and errors tied to their input. skill · 3,870 words · MIT Read the result

Frontend designMeasured, earlier harness Interface state audit Enumerates every state a screen can actually be in, decides which are reachable, and specifies what each must show. The taxonomy is the asset. skill · 2,543 words · MIT Read the result

SEOMeasured Pre-publish page audit Audits a page in order of consequence rather than in order of the checklist, so whether the page can be indexed at all is settled before anyone discusses the title tag. skill · 2,408 words · MIT Read the result

99 skills we have not measured yet

Written, reviewed and free to take. No run behind them, so no claim about what they do to an output. What that means.

Frontend designNot yet measured Accessible name and role audit Audits what a control is actually called and what it claims to be, using the name computation order and the accessibility tree rather than a reading of the markup. skill · 3,998 words · MIT Read the write-up

Product managementNot yet measured Activation and first-run design Activation is a behaviour that predicts retention, not a screen. The test deciding whether a candidate qualifies is empirical, and the honest branch is marking it provisional. skill · 3,938 words · MIT Read the write-up

MarketingNot yet measured Affiliate programme launch kit Produces the programme rather than reviewing one: a payout day computed from the refund window, the clauses written before recruitment, and a kit of finished assets. skill · 5,208 words · MIT Read the write-up

Humanize a draftNot yet measured AI disclosure audit A structural map of disclosure obligations, a taxonomy of seven kinds of use that makes the question decidable, and the list of outputs that must never ship at all. skill · 3,259 words · MIT Read the write-up

SEONot yet measured AI search visibility audit Treats being cited by an answer engine as a different job from ranking, and is blunt about the fact that nobody can currently measure the result reliably. skill · 3,405 words · MIT Read the write-up

Humanize a draftNot yet measured AI tell removal, the word list Forty-eight banned words and phrases, seven construction patterns, and a punctuation count that is invalidated by any later edit, so it runs last and runs again. skill · 3,518 words · MIT Read the write-up

ResearchNot yet measured Analysis sanity check The named reversals that make a confident chart say the opposite of what it appears to, each with the specific check that catches it before anyone presents it. skill · 3,101 words · MIT Read the write-up

SEONot yet measured Answer engine positioning brief Produces a reusable source file rather than a critique: frozen strings for the category label and the positioning sentence, questions written as typed, and a matrix in which no cell may say yes. skill · 3,416 words · MIT Read the write-up

Code reviewNot yet measured API breaking change review Removing a field is the one everybody catches. The dangerous changes keep the field, keep the type, and change what it means. skill · 3,396 words · MIT Read the write-up

Writing and editingNot yet measured Article body spec Produces the production spec rather than checking a draft against one: element order, counts per element, a word ladder with a reason at both ends, and the markup rules the pipeline depends on. skill · 4,178 words · MIT Read the write-up

Writing and editingNot yet measured Article visual set Produces the visual set rather than judging one: two visuals in two placement bands, an infographic bound word for word to the article's table, and nothing readable baked into a raster. skill · 4,908 words · MIT Read the write-up

CopywritingNot yet measured Audience theme pack Generates the theme, the audience list and five example artifacts for one narrow segment, with a specificity bar and a structural-difference check that stops the five collapsing into one. skill · 4,195 words · MIT Read the write-up

CopywritingNot yet measured Broadcast calendar build Produces the calendar, the two behavioural segment definitions and the per-campaign cadence, then ranks twelve offer mechanics on revenue per recipient rather than on opens. skill · 3,487 words · MIT Read the write-up

MarketingNot yet measured Capture surface set Capture is a set, not a widget. This assigns one surface and one trigger to each page type, kills exit-intent on mobile for a reason, and measures each surface on its own. skill · 3,826 words · MIT Read the write-up

PresentationsNot yet measured Chart honesty audit Audits every chart in a finished deck against the way people actually read encodings, and returns a per-chart verdict rather than a list of styling preferences. skill · 3,497 words · MIT Read the write-up

Humanize a draftNot yet measured Claim and hedge audit Audits a draft claim by claim: which hedges carry real uncertainty, which are padding, which numbers have no source, and which claims cannot be wrong. skill · 3,061 words · MIT Read the write-up

ResearchNot yet measured Claim provenance trace Traces a number to the earliest document that actually contains it, names what changed at each hop, and treats an empty origin as the finding rather than a dead end. skill · 2,716 words · MIT Read the write-up

SEONot yet measured Cluster topology blueprint Produces the map rather than the schedule: one greppable pillar declaration per cluster, a closure pre-flight that stops the run, posture and saturation labels, and per-row ownership walls. skill · 3,255 words · MIT Read the write-up

CopywritingNot yet measured Cold outreach sequence Produces the touches and every reply branch for one-to-one outreach, with a research gate that decides whether you may send at all and a rejection branch that stops selling. skill · 3,768 words · MIT Read the write-up

MarketingNot yet measured Comment-led growth system First-hour engagement comes from the followers you already have, so this checks whether they are the right ones before a single comment gets written. skill · 3,100 words · MIT Read the write-up

ResearchNot yet measured Competitive teardown Ranks every claim in a competitive comparison by how it was obtained, and requires each cell to carry a number, a plan name or a named capability rather than a tick. skill · 3,098 words · MIT Read the write-up

Code reviewNot yet measured Concurrency and race condition review Walks seven named race shapes over a diff instead of warning generally about threads, and states which of the four fixes each one takes. skill · 3,849 words · MIT Read the write-up

Build MCP serversNot yet measured Connector directory submission The live policy and terms, the newly prohibited categories, the tool-contract rules that get a server rejected on sight, and what the submission portal actually asks for. skill · 3,518 words · MIT Read the write-up

SEONot yet measured Content cannibalisation audit Detects overlapping pages before rankings do, from titles, shared competitor URLs and token overlap, then returns at most five ranked actions and changes nothing itself. skill · 2,527 words · MIT Read the write-up

SEONot yet measured Content publish run Runs the publishing job rather than reviewing it: backfill-first selection, reconcile-before-write, reciprocal links before translations, and a translated slug that must equal its parent. skill · 4,123 words · MIT Read the write-up

SEONot yet measured Crawl control audit Separates crawl control from index control, which are different mechanisms that cancel each other out whenever somebody uses one to do the other's job. skill · 3,701 words · MIT Read the write-up

PresentationsNot yet measured Data story build Dataset in, chart sequence out. One claim per chart, the claim type selects the encoding, and the axis rules arrive as construction constraints rather than review findings. It draws nothing. skill · 3,507 words · MIT Read the write-up

Frontend designNot yet measured Data table design review Names the three narrow-width patterns and what each one costs, the sorting defects that make a table feel broken, and the four states a table needs rather than one. skill · 3,974 words · MIT Read the write-up

PresentationsNot yet measured Decision memo build Produces the one-pager that survives being forwarded: recommendation first, a fixed slot order, a body and appendix rule, and three shapes for when no recommendation is possible. skill · 3,388 words · MIT Read the write-up

PresentationsNot yet measured Deck accessibility audit Reviews a deck that already exists for the defects assistive technology hits and sighted review never sees, starting with the per-slide object order. skill · 3,719 words · MIT Read the write-up

PresentationsNot yet measured Deck generation constraints The unit and layout detail that decides whether a programmatically produced deck opens correctly on somebody else's machine. It does not generate a deck. skill · 4,250 words · MIT Read the write-up

PresentationsNot yet measured Deck narrative audit It does not build a deck. It reads a finished one and checks whether the headlines, alone and in order, form something a reader could follow. skill · 3,016 words · MIT Read the write-up

Code reviewNot yet measured Dependency upgrade review A patch release can change a default, and a one-line manifest bump can move fifty packages. The review is of the lockfile and the publisher, not the number. skill · 3,538 words · MIT Read the write-up

Frontend designNot yet measured Design system conformant UI Builds a new screen out of what already exists, escalates to a new component or a new package only with a written reason, and checks the result by search rather than by taste. skill · 3,411 words · MIT Read the write-up

Frontend designNot yet measured Design token and theming audit Audits the structure of a token set rather than its taste: which layer a component is allowed to reference, and whether every semantic token resolves in all three theme states. skill · 4,282 words · MIT Read the write-up

Writing and editingNot yet measured Documentation architecture Classifies documentation by what the reader is doing while reading it, and looks for the page trying to be a tutorial and a reference at once, before anyone argues about its prose. skill · 3,592 words · MIT Read the write-up

Writing and editingNot yet measured Draft hold and re-angle Resolves the collision rather than detecting it: quarantine in three places, one verdict from a closed set of five, and a re-angle that narrows to a new sub-intent and links up to the live page. skill · 3,083 words · MIT Read the write-up

Writing and editingNot yet measured Error message writing A message that only says what happened is a notification, not an error message. The missing parts are why it happened and what to do next. skill · 3,575 words · MIT Read the write-up

ResearchNot yet measured Evidence synthesis build Turns a pile of sources into one synthesis with the conflicts classified by cause, weighted on independence rather than reputation, and a required section on what would overturn it. skill · 3,247 words · MIT Read the write-up

Product managementNot yet measured Expert council debate A structured debate method where the lenses are disciplines and published frameworks rather than people, divergence is the output, and the synthesis has to commit to one path anyway. skill · 3,574 words · MIT Read the write-up

Frontend designNot yet measured Focus and keyboard audit Audits a screen for keyboard operability in the order defects actually bite: reachability, then focus order, then focus placement after the DOM changes under the user. skill · 4,452 words · MIT Read the write-up

SEONot yet measured Free tool SEO play Produces a tool concept, a validated keyword cluster and a build specification, or a documented refusal, which is the more common outcome and an equally correct one. skill · 3,444 words · MIT Read the write-up

MarketingNot yet measured Giveaway campaign design The screen immediately after entry decides the campaign. This designs the prize, the share-first entry flow, the partner sequence and the media kit around a coefficient above one. skill · 5,158 words · MIT Read the write-up

MarketingNot yet measured Growth loop audit Most things called loops are funnels drawn in a circle. This decides which yours is, then computes the branching factor and the cycle time that everyone omits. skill · 4,157 words · MIT Read the write-up

Writing and editingNot yet measured Headline and title tests A stranger who has read only the title should be able to say what the piece is about, and the title should name something that exists and can be pointed at. skill · 2,655 words · MIT Read the write-up

Humanize a draftNot yet measured House style guide build Derives rules from a real corpus instead of importing someone else's guide, separates the checkable rules from the judgement calls, and ships every rule with both examples. skill · 3,165 words · MIT Read the write-up

Product managementNot yet measured Idea to scaffolded product Runs the unglamorous scaffolding of a new product end to end, and its real content is the autonomy model: where it stops, where it refuses to guess, and how it stays legible while running. skill · 3,567 words · MIT Read the write-up

SEONot yet measured Internal link graph audit Assesses internal linking as a graph: orphans, dead ends, depth and broken pillar pairs, from the parsed body rather than the stored link field that is usually empty. skill · 3,181 words · MIT Read the write-up

ResearchNot yet measured Interview guide build Produces a phased interview guide in which every question can be answered wrongly, with the product mention placed after the evidence phases rather than in the middle of them. skill · 3,178 words · MIT Read the write-up

SEONot yet measured Keyword portfolio build Turns a product description into a 50-row keyword portfolio with a fixed composition, numeric thresholds per bucket, and a rejection rule for the terms that look fine and are not. skill · 2,876 words · MIT Read the write-up

MarketingNot yet measured Launch plan audit Audits a launch plan for the two ordering errors that sink most of them: long-lead items started too late, and one-shot moves spent before they can convert. skill · 3,840 words · MIT Read the write-up

MarketingNot yet measured Launch runway plan Produces the launch plan rather than reviewing one: two dated goals, a channel table that sums, twelve dated weeks, and the five post-launch pillars most plans omit. skill · 4,115 words · MIT Read the write-up

SEONot yet measured Link target registry Produces the curated allowlist that makes a correct internal link possible: confirmed pages, pages to verify first, an explicit not-real list, and a routing column that picks the target. skill · 3,240 words · MIT Read the write-up

MarketingNot yet measured LinkedIn post audit Checks the first two lines against where the feed actually cuts, then formatting, links, documents, hooks, and the patterns that make a draft read as machine written. skill · 3,568 words · MIT Read the write-up

ResearchNot yet measured Literature search strategy Turns a query into a designed search with a written protocol that can be rerun, plus a scaled-down two-hour version for a commercial question that is not a review. skill · 3,440 words · MIT Read the write-up

Build MCP serversNot yet measured MCP connector build The end-to-end build sequence for an MCP server on revision 2026-07-28, pinned to verified package versions, handing off at each seam to the five skills that own the hard parts. skill · 3,114 words · MIT Read the write-up

Build MCP serversNot yet measured MCP server OAuth setup Names its protocol revision first, then walks the resource server build order: metadata document, canonical resource URI, 401 challenge, audience validation, client registration, issuer check. skill · 4,220 words · MIT Read the write-up

Build MCP serversNot yet measured MCP server test harness Three verification layers with the assertions each one can and cannot make, plus the Inspector CLI exit-code taxonomy and the flag that stops CI opening a browser. skill · 3,143 words · MIT Read the write-up

Build MCP serversNot yet measured MCP tool schema design The field set, the two conflicting name limits, the annotation defaults that work against you, and the x-mcp-header rules whose only symptom is the tool vanishing from tools/list. skill · 4,142 words · MIT Read the write-up

Code reviewNot yet measured Database migration review Classifies every statement in a migration as metadata-only, scanning, or rewriting, then checks the deploy order against the code that depends on it. skill · 3,421 words · MIT Read the write-up

PresentationsNot yet measured Narrative spine build Produces a spine document rather than slides. The ask is written first, the context last, and the sequence is tested by reading the claims aloud with nothing between them. skill · 3,395 words · MIT Read the write-up

MarketingNot yet measured Offer construction audit Separates the offer from the product, then engineers the risk clause: guarantee type, what each costs to honour, who adjudicates it, and which buyers each one attracts. skill · 3,031 words · MIT Read the write-up

CopywritingNot yet measured Offer page build, the fifteen slots Fifteen slots in order, the seven questions answered before drafting, and the placement rules for the damaging admission, the mandatory P.S. and the who-it-is-not-for line. skill · 3,741 words · MIT Read the write-up

MarketingNot yet measured Paid acquisition campaign build Builds the campaign rather than reviewing one: readiness threshold, event set, a three-creative test on one cold audience, a winner inside a stated spend band, then scaling by audience. skill · 4,306 words · MIT Read the write-up

MarketingNot yet measured Paid acquisition readiness gate Runs the gates that decide whether paid traffic can work at all, before any of it is bought, because most paid failures are settled by arithmetic rather than by execution. skill · 4,382 words · MIT Read the write-up

CopywritingNot yet measured Pain-first social posts A hook rule that bans simile and metaphor from line one, four ways to bridge to the product as relief, and six structural patterns for producing variants that differ. skill · 3,654 words · MIT Read the write-up

Frontend designNot yet measured Perceived performance audit Decides what a wait should look like based on how long it is, then checks the specific front-end causes behind LCP, CLS and INP rather than chasing a lab score. skill · 3,420 words · MIT Read the write-up

Product managementNot yet measured Persona set build Demographics predict nothing about a product decision and constraints predict most of it, so the cards carry budget ceilings and incumbents, and the disagreements between them are the finding. skill · 2,876 words · MIT Read the write-up

Writing and editingNot yet measured Plain language rewrite Plain language is a set of techniques with a published standard behind it, not simplification, and the plain version is sometimes longer than the original. skill · 3,381 words · MIT Read the write-up

MarketingNot yet measured Positioning audit Rebuilds a positioning statement from its five inputs, starting with the competitive alternative buyers would really pick, which is usually a spreadsheet or nothing at all. skill · 2,990 words · MIT Read the write-up

MarketingNot yet measured Positioning gate spec Turns a decided position into an enforceable checklist: backlog rows are born tagged with an angle, priority rows must cite two differentiators as mechanisms, and a second failure defers the row. skill · 3,052 words · MIT Read the write-up

MarketingNot yet measured Pricing and packaging audit Starts with the value metric, because per-seat pricing on a product that removes seats charges less the better it works, and no amount of tier design recovers from that. skill · 3,805 words · MIT Read the write-up

Product managementNot yet measured Pricing and packaging build The routing tree is ours rather than received wisdom, and the file says so on its first page: four symptoms, four remedy classes, and one branch whose instruction is not to touch the price. skill · 3,761 words · MIT Read the write-up

MarketingNot yet measured Product truth pack One numbered file per feature, each ending in honest limits, plus a register of exact banned strings scoped to alt text and words rendered inside images, every ban carrying its replacement. skill · 3,351 words · MIT Read the write-up

SEONot yet measured Published corpus mirror Produces the corpus every later check reads: one file per article per language, a derived index regenerated rather than edited, and a second snapshot that proves the mirror is not fiction. skill · 3,356 words · MIT Read the write-up

CopywritingNot yet measured Release announcement set Four surfaces, one release. The changelog is the only one written for someone searching, which is why it is the only one that has to state the old behaviour as well as the new. skill · 4,441 words · MIT Read the write-up

Build MCP serversNot yet measured Remote MCP deploy Owns the transport conformance and the box it runs in: required headers, the status and error code pairings, SSE buffering and cancellation, horizontal scaling, and era routing. skill · 3,790 words · MIT Read the write-up

MarketingNot yet measured Review harvest campaign Produces the campaign rather than auditing one: two timing windows, a story prompt instead of a review request, a page instead of instructions, and the incentive rule that keeps the reviews up. skill · 4,523 words · MIT Read the write-up

Humanize a draftNot yet measured Sentence rhythm edit Counts words per sentence and reads the spread, not the mean, then fixes the five rhythms that flatten a passage: repeated lengths, repeated openers, and trailing qualifiers. skill · 2,978 words · MIT Read the write-up

Writing and editingNot yet measured Series continuity ledger Produces the canon file a bylined series is written against: every number stated in print, standing corrections that must not be reverted, and a shared log that makes the promotion rule enforceable. skill · 3,166 words · MIT Read the write-up

SEONot yet measured SERP reverse-engineered article Produces the article from a captured results page rather than from a topic, and refuses to publish a draft that is merely as good as the pages it is trying to outrank. skill · 2,799 words · MIT Read the write-up

SEONot yet measured Site migration audit Treats release order and the completeness of the redirect map as the whole job, then separates a normal post-launch dip from a migration that is actually broken. skill · 4,290 words · MIT Read the write-up

PresentationsNot yet measured Slide density audit Scores each slide of a finished deck for density and legibility, then hands back a ranked list of the worst offenders rather than a comment on all forty. skill · 3,217 words · MIT Read the write-up

ResearchNot yet measured Source credibility triage Rates a source and says why, with a hierarchy that inverts for specification claims and a cannot-assess verdict that counts as a real answer rather than a failure. skill · 3,371 words · MIT Read the write-up

PresentationsNot yet measured Speaker notes and timing audit Audits the notes and the timing of a deck that already exists, on the rule that the slide, the spoken word and the notes each carry a different thing. skill · 3,297 words · MIT Read the write-up

Product managementNot yet measured Spec to tickets Writes the specification and the tickets, but resolves every field, endpoint and table against the real codebase before any of them reach the document. skill · 3,438 words · MIT Read the write-up

Humanize a draftNot yet measured Structure de-templating Eight named template shapes, a diagnostic that reads only the headings, a section-length distribution check, and the four openings that work. Shape survives every word-level edit. skill · 3,191 words · MIT Read the write-up

Writing and editingNot yet measured Style guide conformance Separates mechanical findings from judgement ones so the human queue stays short, and treats a synonym introduced for variety as a defect rather than as good writing. skill · 3,426 words · MIT Read the write-up

ResearchNot yet measured Survey design audit Audits a questionnaire while the damage is still cheap to undo, and treats who never answered, rather than how many did, as the dominant threat to the result. skill · 3,273 words · MIT Read the write-up

Code reviewNot yet measured Test suite review A named taxonomy of bad tests, each with the tell that identifies it, plus a scoring procedure a reviewer can run over a suite in an afternoon. skill · 3,476 words · MIT Read the write-up

SEONot yet measured Topic cluster backlog Turns a keyword portfolio into a 30-row schedule with a fixed type quota, word counts bound to measured demand, and link targets that already exist by the time the row is written. skill · 3,552 words · MIT Read the write-up

SEONot yet measured Traffic drop forensics Diagnoses a traffic fall in the order that stops you fixing the wrong thing, and catalogues the failures that are invisible to anyone looking at the rendered page. skill · 3,826 words · MIT Read the write-up

Writing and editingNot yet measured Tutorial reproducibility audit Reads a tutorial as a reader who has nothing installed, on Mac and on Windows, and treats every step with no observable check as a failure that will surface later and cost more. skill · 3,455 words · MIT Read the write-up

Product managementNot yet measured Unblock decision protocol Most autonomous systems stop when any single condition looks risky, which is why they stall. Here a halt needs all four at once, and everything else becomes a decision plus a deferred item. skill · 3,155 words · MIT Read the write-up

Product managementNot yet measured Viability blueprint A structured pre-launch interview that produces a go-to-market blueprint, built around the rule that you ask one to three questions at a time and never accept a vague answer. skill · 3,809 words · MIT Read the write-up

Humanize a draftNot yet measured Voice match from a corpus Extracts a countable fingerprint from a corpus of the author's real writing, measures the draft on the same features, then closes the gaps in order of size. skill · 3,247 words · MIT Read the write-up

Frontend designNot yet measured WCAG 2.2 audit Audits a page against WCAG 2.2 level AA by test procedure rather than by criterion number, and separates what an automated scan can settle from what it cannot. skill · 5,111 words · MIT Read the write-up

Frontend designNot yet measured Interactive 3D on the web, review and budget Budgets an interactive 3D scene against the page it is going into, then decides whether it should exist at all, with the arithmetic written down rather than guessed. skill · 3,930 words · MIT Read the write-up

Frontend designNot yet measured Interactive 3D scene embed Builds the integration rather than judging one. Qualifies the scene, sets the budgets before it is authored, and loads it so the page still works when it is deleted. skill · 4,143 words · MIT Read the write-up

By task

Browse by task

Skills are organised by the job you are doing, not the job title you have.

Build MCP servers Six skills for the build side: the server itself, the transport and hosting, the authorisation flow, the tool schema, the test harness, and the submission package. 6 skills · 0 measured Code review Review work where the knowledge is the hard part: which lock a migration takes, which shapes a race can have, what actually breaks a consumer of your API, and where the rest of the family of this bug lives. 6 skills · 1 measured Copywriting Six skills that produce copy rather than grading it. Each one carries a slot order, a device bank or a set of placement rules, because the model already writes a decent sentence and the ordering is where the work actually is. 6 skills · 0 measured Frontend design The review side of front end work: the states nobody built, the attributes that decide whether a form autofills, the contrast maths, and where focus goes when the focused element is deleted. 11 skills · 2 measured Humanize a draft Six separate passes, because the tells that give generated prose away live at six different levels and a single rewrite instruction reaches none of them reliably. 7 skills · 0 measured Marketing The parts of marketing that have a factual core a model gets wrong: authentication records, platform limits, named legal obligations, and the order operations have to happen in. 17 skills · 1 measured Presentations The deck already exists. These are for deciding whether it says one thing per slide, whether the chart is honest, and whether it survives being read rather than presented. 9 skills · 0 measured Product management Six skills that produce the specification, the tickets, the blueprint and the decision. Each carries an ordering, a gate or a template where getting the sequence wrong is the expensive part. 8 skills · 0 measured Research Where a number came from, whether the source is what it appears to be, and the named biases that make a confident analysis wrong. 8 skills · 0 measured SEO Search work with a checkable core: what a results page is actually made of, which of your own pages are competing with each other, and why the traffic went away. 16 skills · 1 measured Writing and editing Editing as an ordered set of passes, documentation as a typed artefact, and the two headline tests that replace taste with something falsifiable. 10 skills · 0 measured

The evidence

Why some cards say measured and most do not

Measured. 5 of them. Each was given a realistic task on real material, then the identical task was run again with the skill removed, 5 runs each way. Every output was graded alone against a rubric written by someone who had never seen the skill, by a session that was not told the other arm existed. Whether it passed was decided by a rule written down before a single run executed. Those pages carry the worst case, the median, the p value, and what the skill costs you as well as what it buys.

Not yet measured. 99 of them. Written, read for accuracy, free to take, and we have run no experiment on them, so we make no claim about what they do to an output. Their cards say so in the grid and their pages say so in a grey box above everything else. Each one names the test it is waiting for.

Cut. 4 of them, and you cannot install those, because they are not published. A skill that went through the harness and failed does not get a quiet demotion to the unmeasured pile. It is deleted, and its numbers stay in the results table permanently.

Unmeasured is not the same as failed. Everything in the second group might turn out to work, might turn out to do nothing, and we will tell you which when we have run it. The full method, the results table for every skill we ever tested, and the round we had to throw away.

Free to copy. MIT. Mac and Windows.

One click, or no click at all

Import a skill straight into your setup.

Locul puts a skill in the right folder for Claude Code, Claude Desktop and Cursor, scrubs any environment variables before they reach you, and keeps the memory those skills read from up to date by itself. Mac and Windows.

Start free