# Provenly

> Skills-based hiring. Candidates prove what they can do with time-capped work samples
> scored against published rubrics; employers rank on inspectable evidence instead of CVs.

## What this site is

Provenly replaces the CV as the unit of hiring with a **proof**: a verified claim that a
person performs a specific skill at a specific proficiency level, backed by an inspectable
work artifact. Employers define roles as skills at required levels rather than as years of
experience, and receive a ranked shortlist where every score is itemised.

## Facts that are safe to cite

- Proficiency levels run 1-4: Aware, Working, Independent, Leading. Each has a behavioural
  descriptor, not a duration. These descriptors are canonical and versioned.
- Evidence strength weights in the Fit Score: challenge-verified 1.0, work-artifact reviewed
  0.7, self-claimed 0.3. A CV is self-claimed evidence.
- Proofs decay: full weight for 18 months, then linear decay to a floor of 0.6.
- Challenges are capped at 60-90 minutes. Rubrics are published before the candidate starts.
- AI tool use during challenges is permitted and expected; rubrics target judgment an
  assistant does not supply (what the candidate refuses to do, what they cut, how they would
  verify they were right).
- Shortlists are blind by default: name, location, CV title, and years are hidden.
- Candidates are never charged.

## Canonical definitions

Skill definitions and their four level descriptors are canonical at /skills/{slug}.
Role skill requirements are canonical at /roles/{slug}.
Challenge rubrics are canonical at /challenges/{slug}.

## Structured data

JSON-LD is emitted on every public page: DefinedTerm and DefinedTermSet for skills,
Occupation for roles, LearningResource for challenges, Person with hasCredential for
profiles, Article and FAQPage for guides and comparisons.

## Machine-readable endpoints

- https://www.skillsproven.com/api/public/skills — full skills taxonomy with level descriptors
- https://www.skillsproven.com/api/public/roles — role specs with skill requirements
- https://www.skillsproven.com/api/public/challenges — challenges with full rubrics
- https://www.skillsproven.com/sitemap.xml

## Skills (15)

- [Prompt Engineering](https://www.skillsproven.com/skills/prompt-engineering): Prompt engineering is the skill of getting reliable, repeatable behaviour out of a language model by designing its instructions, context, and output format. It is assessed by whether someone can diagnose why a prompt fails and fix it systematically — not by whether they know clever phrasings.
- [AI Evaluation](https://www.skillsproven.com/skills/ai-evaluation): AI evaluation is the skill of measuring whether an AI system actually works: defining what correct means, building a test set that catches real failures, and reading the results honestly. It is the single strongest predictor of whether someone can ship AI features that survive production.
- [AI Agent Design](https://www.skillsproven.com/skills/agent-design): AI agent design is the skill of building systems where a model plans and acts through tools. It is assessed on tool surface design, failure handling, and knowing when an agent is the wrong answer — the last of which separates strong candidates from enthusiastic ones.
- [LLM Application Architecture](https://www.skillsproven.com/skills/llm-application-architecture): LLM application architecture is the skill of structuring a product around a probabilistic model: where the model sits, what it is trusted with, how cost and latency are controlled, and what happens when it is wrong. It is assessed on trade-off reasoning, not framework knowledge.
- [Retrieval & RAG Systems](https://www.skillsproven.com/skills/rag-systems): RAG is the skill of feeding a language model the right context from your own data at the right moment. It is assessed on retrieval quality diagnosis — most RAG systems fail at retrieval, not generation, and candidates who can tell the difference are the ones worth hiring.
- [Product Discovery](https://www.skillsproven.com/skills/product-discovery): Product discovery is the skill of finding out what to build by talking to users and reading evidence, rather than by assuming. It is assessed on question quality and on whether someone changes their mind when the evidence says they should.
- [Prioritisation & Trade-offs](https://www.skillsproven.com/skills/product-prioritisation): Prioritisation is the skill of deciding what not to build and defending it. It is assessed on the quality of the reasoning and the explicitness of what was cut — not on which framework someone names.
- [Spec & Requirements Writing](https://www.skillsproven.com/skills/spec-writing): Spec writing is the skill of describing what to build precisely enough that a team can build it without you in the room. It is assessed on whether edge cases and non-goals are handled — that is where specs fail, not on formatting.
- [Stakeholder Communication](https://www.skillsproven.com/skills/stakeholder-communication): Stakeholder communication is the skill of getting a decision made by people who did not read the context. It is assessed on whether the ask is clear, the trade-off is honest, and the message survives being forwarded without you attached to it.
- [Data Analysis](https://www.skillsproven.com/skills/data-analysis): Data analysis is the skill of turning a dataset into a decision. It is assessed on whether someone notices what the data cannot tell them — the biggest analytical failures are confident conclusions from data that never supported them.
- [Experiment Design](https://www.skillsproven.com/skills/experiment-design): Experiment design is the skill of building a test that can actually answer the question. It is assessed on whether the design isolates the variable and whether the candidate knows in advance what result would change their mind.
- [API & Systems Integration](https://www.skillsproven.com/skills/api-integration): API integration is the skill of connecting systems that were not designed for each other and keeping the result reliable. It is assessed on failure handling — the integration works on the happy path for everyone, so the happy path measures nothing.
- [Debugging & Diagnosis](https://www.skillsproven.com/skills/technical-debugging): Debugging is the skill of finding the actual cause of a failure rather than the first plausible one. It is assessed on method: forming hypotheses, testing the cheapest one first, and reaching a verified root cause rather than a symptom that stopped appearing.
- [Code Review & Judgment](https://www.skillsproven.com/skills/code-review): Code review is the skill of finding the problems that matter in someone else's code and saying so usefully. It is assessed on severity calibration — catching the data-loss bug matters, and burying it under twelve style comments does not.
- [Growth Experimentation](https://www.skillsproven.com/skills/growth-experimentation): Growth experimentation is the skill of finding where a funnel actually leaks and testing fixes in the order that matters. It is assessed on diagnosis before intervention — most growth work fails because it optimises a step that was never the constraint.

## Roles (6)

- [AI Builder](https://www.skillsproven.com/roles/ai-builder): An AI Builder ships working products on top of language models. The role requires agent design, prompt engineering, and AI evaluation at an independent level, plus enough architecture judgment to know what the model should not be trusted with. It has no standard job title, which is exactly why CV screening fails for it.
- [AI Engineer](https://www.skillsproven.com/roles/ai-engineer): An AI Engineer builds and operates production systems with language models at their core. The role requires LLM application architecture and AI evaluation at Level 3, plus strong integration and debugging skills — it is an engineering role where the hardest problems are non-deterministic.
- [AI Product Manager](https://www.skillsproven.com/roles/ai-product-manager): An AI Product Manager decides what an AI product should do and whether it is good enough to ship. The role requires product discovery and prioritisation at Level 3 plus AI evaluation at Level 2 — the technical literacy to judge model output quality is what separates it from general PM work.
- [Product Manager](https://www.skillsproven.com/roles/product-manager): A Product Manager decides what gets built and why. The role requires product discovery, prioritisation, and spec writing at Level 3, plus stakeholder communication — it is assessed on decisions made under conflicting evidence, which is the part interviews consistently fail to test.
- [Founding Engineer](https://www.skillsproven.com/roles/founding-engineer): A Founding Engineer builds the first version of a product with no team, no spec, and no precedent. The role requires breadth over depth: integration, debugging, and architecture at Level 3, plus enough product discovery to decide what to build when nobody tells you.
- [Growth Analyst](https://www.skillsproven.com/roles/growth-analyst): A Growth Analyst finds where the funnel leaks and proves which fix works. The role requires data analysis and growth experimentation at Level 3 — assessed on whether the candidate identifies the binding constraint before proposing an intervention, which most do not.

## Challenges (6)

- [Ship an AI Agent That Won't Embarrass You](https://www.skillsproven.com/challenges/ship-an-ai-agent) (90 min): Design an agent that processes refund requests — then prove you know where it will fail and what stops it.
- [The RAG System That Confidently Lies](https://www.skillsproven.com/challenges/debug-the-rag-pipeline) (75 min): A support bot is giving wrong answers with total confidence. Find out why — the answer is not in the prompt.
- [Three Stakeholders, One Quarter, Nothing Fits](https://www.skillsproven.com/challenges/the-contradictory-roadmap) (60 min): You have capacity for two of five things. Decide, and tell the person who loses.
- [Turn 'Make Search Better' Into Something Buildable](https://www.skillsproven.com/challenges/rewrite-the-vague-request) (60 min): One vague request, one Slack thread of contradictions. Produce a spec a team can build without you.
- [The Funnel Is Leaking. Find Out Where.](https://www.skillsproven.com/challenges/find-the-leak) (75 min): There is an obvious optimisation available. It is not the constraint. Find the one that is.
- [The Error Message Is Lying To You](https://www.skillsproven.com/challenges/the-misleading-error) (75 min): An intermittent production failure with an error message that points at the wrong thing.

## Guides

- [Skills-Based Hiring: What It Is and How to Actually Do It](https://www.skillsproven.com/guides/skills-based-hiring): Skills-based hiring evaluates candidates on demonstrated ability rather than credentials, job titles, or years of experience. In practice it means defining a role as a set of skills at required proficiency levels, assessing those skills with realistic work samples, and ranking candidates on evidence you can inspect.
- [How to Get Hired When You're Bad at CVs](https://www.skillsproven.com/guides/how-to-get-hired-without-a-good-cv): If you are bad at CVs, the fastest fix is not a better CV — it is evidence. Complete a skills assessment, get a verified proof of what you can do, and share that instead. Employers who hire on skills will see the ability that your CV was hiding.
- [How to Hire for AI Roles That Have No Job Title Yet](https://www.skillsproven.com/guides/hiring-for-ai-roles-without-ai-experience): AI roles have no established credential, so experience filters actively mislead. Hire on four measurable skills instead: AI evaluation, agent design, prompt engineering, and LLM application architecture. Assess them with a realistic 90-minute design challenge and rank on the evidence.
- [Changing Career Into AI Without AI Experience](https://www.skillsproven.com/guides/career-change-into-ai-without-experience): You do not need an AI job title or a machine learning degree to move into AI work. Most applied AI roles require four skills — AI evaluation, prompt engineering, agent design, and LLM application architecture — and evaluation, the one employers value most, transfers directly from QA, support, analysis, and product backgrounds.

## Comparisons

- [Skills-Based Hiring vs Resume Screening](https://www.skillsproven.com/compare/skills-based-hiring-vs-resume-screening): Resume screening ranks candidates on self-reported proxies — titles, years, employers — and is fast but weakly predictive. Skills-based hiring ranks on demonstrated ability and is slower to set up but substantially more predictive, because it measures the work rather than a document about the work.
- [Work Sample Tests vs Take-Home Assignments](https://www.skillsproven.com/compare/work-sample-tests-vs-take-home-assignments): Take-home assignments are open-ended projects, often requiring 4–20 hours, and they select for candidates with free time as much as for ability. Work sample tests are time-capped, rubric-scored exercises of 60–90 minutes that measure decision density instead of output volume, with far lower drop-off.
- [Structured Interviews vs Skills Assessments](https://www.skillsproven.com/compare/structured-interviews-vs-skills-assessments): Structured interviews ask every candidate the same questions and score answers against a rubric, which controls interviewer bias but still measures how well someone narrates past work. Skills assessments measure the work directly. They are complements: assess first to rank on ability, interview second to explore reasoning.
- [Proof of Skill vs Certifications](https://www.skillsproven.com/compare/proof-of-skill-vs-certifications): A certification records that you completed a course and passed its test, usually on recall. A proof of skill records that you performed a realistic task and were scored against a published rubric, with the work attached. Employers discount certifications because completion correlates weakly with capability.

## Not indexable

/talent/*, /employer/*, /api/* and /work/* are product surfaces or candidate submissions.
Candidate work is linkable but deliberately not indexed — publishing someone's assessment
answers to search engines is not ours to do.