Coding AI-skills evidence is a report tile on a coding assessment that shows which AI collaboration skills a candidate used while working with an in-platform coding agent — for example Codility’s AI Insights listing five skills with word-for-word quotes from the candidate’s own prompts. Proof they can do the job is a role-specific task whose output a hiring manager can open, read, and defend in a debrief (a work sample or job simulation). Quoting how someone prompted Codex or Claude Code answers “how did they use AI on this coding task?” It does not by itself prove “this person can produce this job’s day-one deliverable on our JD.”
That shelf just got a named product moment. Codility Lab (Christopher Greco, 19 Oct 2026) announced that candidates can use OpenAI’s Codex next to Anthropic’s Claude Code in Codility Screen, Interview, and Skills Intelligence — no account or API key to set up. Each task report’s AI Insights tile lists which of 5 AI skills the candidate showed: Context Engineering, AI Iteration, AI Output Validation, AI Task Delegation, and AI Failure Modes. Every quote must match the transcript word for word; Codility says each skill is judged against a fixed list of 10 to 13 behaviors, the method was tested on 300 real sessions, the tile appears 10 to 15 minutes after the task ends, and the evidence never changes the score. Schmidt and Hunter (Psychological Bulletin, 1998) put work-sample tests near .54 predictive validity for job performance.
This guide is for TA leaders and hiring managers who hear “we now have AI-skills evidence with prompt quotes” and wonder whether a coding-platform report tile replaces openable proof. It separates a vendor AI-fluency evidence layer from hire-time work evidence, and shows where Dotof AI Role-Simulation Assessments fit: after (or beside) the coding AI Insights tile, prove day-one work before you spend scarce panel time. Related (not rewrites): Google AI-fluency interviews, AI fluency in interviews, AI interviewer vs job simulation, and hiring companies.
What does Codility AI-skills evidence actually deliver?
It delivers a reviewer-facing summary of how a candidate worked with an in-platform coding agent on a Codility task. Reports already showed AI usage volume and kept the full transcript; Greco’s 19 Oct 2026 post says judging quality meant reading hundreds of messages, so two candidates with the same score and the same AI usage looked identical. AI Insights now surfaces which of five skills appeared, with one sentence on why each quote counts, across Codex, Claude Code, the AI Copilot, and the AI Assistant. Interviewers can also give Codex instructions the candidate never sees — such as adding a subtle bug — and watch whether the candidate reviews the output.
Treat that as a coding-assessment AI-fluency report tile. Celebrate when quotes make review faster and when the strongest moments are candidates catching the agent out (wrong failing-test counts, wrong expected values). Do not import “they showed AI Output Validation with three quoted prompts” into a new-hire debrief as proof that Candidate A can produce this job’s day-one deliverable on your JD — especially for PM, analyst, sales, ops, or other non-coding roles, and even for SWE roles whose day-one work is not “solve this Codility task with Codex.” The artifacts are a five-skill tile, verbatim prompt quotes, and a coding transcript. They are not a memo, a model, a plan, or a work product written against your live job description.
Why isn’t coding AI-skills evidence enough at hire time?
Because the hiring manager’s job in a debrief is to open evidence of work on this req — not a story about how well someone steered Claude Code on a vendor coding pad. AI Insights answers “how did they use AI on this assessment?” A work sample answers “what would this person produce next week on our JD?” Managers defend shortlists to peers with the second answer. Codility is explicit that the evidence is informational, never changes the score, and that a skill missing from the report is no sign of weakness because many tasks never call for all five — useful honesty that still leaves the hire packet empty of role output.
Do not mix this with neighboring products. A Google AI-fluency interview is an employer-run Gemini code-comprehension bar. Generic AI fluency in interviews is talk-about-tools. An AI interviewer (including HackerRank Chakra-style canvases) is conversational or IDE Q&A at scale. Codility’s moment is a named coding-assessment SKU: Codex beside Claude Code plus an AI Insights tile with five skills and word-for-word prompt quotes. It is not a task generated from your job description for every role you hire, and it is not a claim that Codility buys or uses Dotof.
When should HR use a work sample instead of (or after) coding AI-skills evidence?
- Use AI Insights for coding AI-tool fit when the role truly requires day-to-day work with coding agents and your team will read quoted prompts in debrief. Keep a named human on advance and reject — a five-skill tile is not an unsupervised hire vote.
- Add proof before panel time once the role has observable day-one deliverables beyond “complete this Codility task with Codex,” a wrong hire is costly, or managers refuse to interview without something they can open. Generate the task from the live JD — see from job description to role simulation.
- Prefer the sample over “they showed AI skills evidence” as the hire gate for professional roles where “can they do the work?” is the question you cannot afford to get wrong — including non-engineering roles where a coding pad does not apply, and SWE roles where the debrief still needs openable job output. Use the tile to decide who can collaborate with coding agents. Use the sample to decide who advances on your JD.
Do not ask a prompt-quote tile to answer “can they do Tuesday’s work on our req?” — that is a task, not a five-skill band. Keep a written rule: no “AI Insights cleared” note should be the only hire vote. The practical gates are in the responsible AI hiring checklist. Finding and matching people is a separate product problem — see AI sourcing and matching — and it does not replace the artifact either.
Which should we buy for skills-based hiring in 2026 — coding AI-skills evidence or a work sample?
Buy the sample for the hire gate — and treat coding AI-skills evidence as an optional engineering-assessment layer that still needs openable proof underneath when the decision is day-one work on your JD. AI Insights is a tool-fit purchase: five skills, word-for-word quotes, score unchanged. A work sample is a selection purchase: openable evidence before panel hours. If your buyer question is “AI coding skills evidence vs job simulation — which should we buy?”, start with the artifact managers will open in a debrief — then decide whether a Codex/Claude Code report tile still earns its place for engineering AI-collaboration signal. That is the answer to “coding AI-skills evidence vs job simulation”: keep or add the vendor tile if AI-assisted coding is the job; prove the work either way. See also AI interviewer vs job simulation, AI screening vs human judgment, and the anatomy of a useful AI assessment report for hiring managers.
How does Dotof fit this decision?
Dotof AI Role-Simulation Assessments start from the job description, generate a realistic task, run it in a supervised browser, and return structured evidence hiring managers can compare. Use Codility-style AI-skills evidence when your pain is “we do not know how they collaborate with coding agents on a timed technical task.” Use the simulation when the decision you cannot afford to get wrong is whether this candidate can do the job — especially when the “evidence” so far was a five-skill tile on someone else’s coding problem set. Coding AI-fluency report tiles and job-task proof can coexist. They should not be confused. Nothing in Codility’s 19 Oct 2026 Lab post means Codility buys, endorses, or uses Dotof.