Resumes lie.
Code doesn't.
GitHub is a developer's most truthful resume — code, commit history, and PR collaboration are things you've done, not things you've said. JobAgent turns repositories into verifiable ability profiles, so companies hire people who've actually built things, and candidates prove themselves with their work.
The traditional hiring pipeline is breaking down in the AI era
Resumes are self-descriptions and interviews can be trained — what truly proves ability is the behavioral trail left behind.
Resumes can be mass-generated
AI can write, polish, and fabricate resumes. When HR screens resumes, they're really screening copywriting ability, not work ability.
Interviews can be gamed with prep
Canned questions and question banks are drilled repeatedly, making interview performance increasingly disconnected from real on-the-job ability.
GitHub is also polluted by AI
One-click template repos, inflated stars, padded commits — surface-level data alone can be just as misleading.
Screening costs keep rising
Resume volume explodes alongside misinformation, driving verification and time costs sharply upward for hiring teams.
Hiring and applying revolve around the same real evidence
GitHub is the bridge in the middle: enterprise job requirements and candidate ability profiles both align on the same "real work trail."
The shared analysis core is live: enter a GitHub/Gitee username and get a verifiable ability profile; job matching, tailored resumes, and application tracking are available, while enterprise SaaS subscriptions and interview-outcome feedback open in later iterations.
B-side · EnterpriseHiring side: from screening resumes to verifying evidence
C-side · CandidateApplying side: GitHub as a living resume
Not just repo analysis — a verifiable job-search loop
From behavioral signals to a tailored resume, every claim is backed by real, clickable, reviewable evidence — never fabricated.
Detects the AI-agent era skill set
A curated taxonomy covers ai agents, MCP, RAG, function calling, LLM serving, and more, normalized from code and GitHub topics — not just Python or Java.
GitHub + Gitee evidence
Analyze both public activity trails in one run with automatic cross-source mirror de-duplication, so developers anywhere speak from the same trustworthy evidence.
Authenticity grades, not one score
A dozen-plus signals — author consistency, star inflation, commit bursts, self-merged PRs, external collaboration — map to four grades (likely/mixed/suspicious/insufficient), each traceable.
Job matching + tailored resumes
An aggregated remote-job pool is ranked against your skills with per-skill reasons; generate a role-tailored zh/en resume in one click (optional AI polish that reorders facts, never invents them).
One-click ATS fill extension
A Chrome extension reads the profile inside Greenhouse, Lever, and other ATS pages to auto-fill skills and experience, cutting repetitive forms.
Read-only MCP surface
Exposes standard read-only MCP tools for AI agents and IDEs — get profile, resolve by subject, search jobs, match profile to job — with keyed access, no raw evidence by default, and no triggered analysis.
We don't just look at GitHub — we verify it
In the AI era, GitHub is full of repos that "look hardworking" but aren't. JobAgent's core capability is making evidence trustworthy.
Detect one-click generated / template-cloned repos
Flag repos with no real development process and excessive code similarity, filtering out "project-looking" shells.
Detect star inflation / anomalous commits
Commit time distribution, contributor structure, star growth curves — anomalous inflation patterns have nowhere to hide.
PR · Issue · real collaboration records
Code merged in real projects and discussions participated in are more persuasive than any self-description.
Four phases, from analysis core to two-sided platform
No exclusive choices — technical and non-technical roles, enterprise and candidate services, overseas and domestic markets, all included. Rolled out in M1→M4 phases, each validated before moving to the next.
Analysis Core + Report
L0/L1 pipeline (API-only, no clone), AbilityProfile v0.1, authenticity grading + evidence, real-project interview questions, landing page Demo connected to real engine.
- Enter GitHub username → verifiable profile report
- Four-level authenticity grading, no single percentage
- 20–50 labeled accounts de-risking experiment
Self-claim + Light B-side
Developers claim own profiles via GitHub/Gitee OAuth, with improvement suggestions, recruiter declaration, and a public concise view. L3 semantic-layer interview enhancement remains planned.
- Self-authorized/shared reports are "authoritative"
- Unauthorized access gets public light preview with clear labeling
- Hiring teams paste public accounts to see concise profiles
Two-sided Loop
Job-pool aggregation, profile-vs-job matching/ranking, job recommendations, tailored resumes, and application/interview tracking are live; one-click apply, interview-outcome feedback, and paid subscription tiers remain planned.
- Live: job recommendations · tailored resumes · application tracking
- Planned: enterprise SaaS hiring subscription
- Outcome label feedback = long-term data moat
Multi-platform / Non-technical
Gitee is integrated with dual-source fusion; L2 clone static analysis (sandboxed), L4 cross-repo identity aggregation, and non-technical multi-evidence sources (portfolios/cases/certificates) remain planned.
- GitHub + Gitee dual-source fusion is live
- Extend to non-technical roles after technical roles work
- Unified EvidenceItem structure for multi-source intake
Enter your GitHub username
Enter a GitHub username to instantly generate a verifiable ability profile — authenticity grading, ability tags, and real-project evidence, based on public activity, no login required.