Frequently asked questions

skill·proofer cross-references a candidate's CV claims against their public GitHub activity and produces a report with three sections: evidence (which claims are backed by code), technical fingerprint (languages, activity, work style), and interview questions (specific to the candidate's actual repos, with expected answers).

It's designed for hiring managers, technical leads, and recruiters who want to move beyond self-reported resumes before investing time in interviews. It's also used by individual developers to audit their own public GitHub profile.

You upload a CV (PDF) and provide a GitHub username. Behind the scenes, a 5-step pipeline runs:

1. CV Parser (AI) — Extracts technical claims from the CV into structured skills.
2. GitHub Fetch — Pulls public repos, commit history, language breakdowns, and READMEs.
3. Evidence Match — Matches each CV claim against repo content. Language claims are checked against byte counts; framework and tool claims are checked against READMEs and descriptions.
4. Fingerprint — Computes primary language, activity level, solo vs. collaborative patterns.
5. Question Gen (AI) — Generates 5 questions per important repo (2 general, 3 code-specific) with what a strong answer would mention.

Steps 3 and 4 are fully deterministic — same input always produces the same output. The whole process takes about 60-90 seconds.

Just two things: the candidate's CV in PDF format and their GitHub username. That's it — no API keys to configure, no setup, no subscription. Each report costs €149 and is ready in about a minute. You can pay per report with no ongoing commitment.

The report will show a No Public Evidence result for claims that can't be verified. This is common and not a negative signal — many experienced developers primarily work in private corporate repositories.

You will still receive the parsed CV claims and any available metadata from limited repos. Interview questions will only be generated if at least one qualifying repository is found (non-fork, more than 5 commits, with a README).

We only access public data — private repositories and private contributions are not visible to us. The tool works best for candidates who have meaningful public activity, even if it's not their primary work.

Each CV claim is matched against repositories using different strategies depending on the type of claim:

Programming languages (e.g., "Python — 4 years") — checked against GitHub's language byte counts across all repos. If a language appears with significant volume across multiple repos, it's marked as Verified.

Frameworks, libraries, and tools (e.g., "React", "Django", "Docker") — checked against README content and repository descriptions. If mentioned in a substantial repo, it's marked as Partial Evidence.

Skill categories (e.g., "Web Scraping — BeautifulSoup, Selenium") — the category is verified if any of its sub-skills are found across repos.

Non-technical claims (leadership, teamwork, etc.) and cloud/BI tools (AWS, Tableau, etc.) are excluded from matching — they either leave no code traces or require proprietary access.

Absence of evidence does not imply absence of skill. A "No Public Evidence" result simply means we couldn't find public GitHub data to support the claim.

AI is used for exactly two tasks: parsing CV text into structured claims and generating interview questions. Both use a dedicated GPT model with a low temperature setting (0.3) for consistent output, and a strict JSON schema to enforce structure.

Everything else — evidence matching, fingerprint computation, and report assembly — is deterministic. The same input always produces the same output, regardless of when or how many times you run it.

For interview questions: the AI works from actual repository code and READMEs, so questions are grounded in real code. Each question includes what a strong answer would mention, giving you a concrete benchmark. However, treat the report as a screening aid, not a definitive judgment. We recommend using it to prepare for interviews, not to make pass/fail decisions.

Yes. Here's exactly what happens with your data:

CV files — Sent to OpenAI for parsing, then stored encrypted in our database. They are not used by OpenAI for model training. You can request deletion at any time.

GitHub data — Fetched via the public GitHub API. Only public information is accessed.

Payment — Processed entirely by Stripe. We never see or store your card details.

Authentication — Handled by Clerk. We only store your email address for account management.

No tracking cookies, no analytics, no third-party marketing. Only essential cookies for authentication. For full details, see our Privacy Policy.

Reports are delivered via the web dashboard and as a downloadable PDF. If you encounter a technical issue (report fails, PDF doesn't generate, data is incorrect), contact us and we will make it right — whether that's regenerating the report or resolving the issue.

Because each report is generated on-demand using paid AI API calls and compute resources unique to each request, we cannot offer refunds for completed reports. If you're unsure whether the tool fits your use case, start by running a report on a candidate with clear public GitHub activity to evaluate the output.