| name | ai-talent-graph |
| description | Academic research-oriented AI scholar profiling tool integrating OpenAlex, arXiv, Semantic Scholar, ORCID free APIs, with optional X/Twitter public influence data. Triggers: search scholar, find expert, AI talent graph, analyze institution, find LLM researchers, latest papers, academic graph, paper tracing, research direction analysis, check scholar twitter, X account influence. Not for: recruitment-oriented analysis (use ai-talent-radar); precise Chinese scholar search (data sources are primarily English). |
ai-talent-graph 7.0.0
First-time Setup
Run the dependency check script before first use:
bash scripts/setup.sh
The agent will auto-run this on first trigger; manual execution is usually unnecessary.
Integrates OpenAlex, arXiv, Semantic Scholar, ORCID multi-source free APIs to build comprehensive AI scholar profiles.
Quick Start
Step 1: Install dependencies
pip install requests
Step 2: Locate script directory
SKILL_DIR=$(dirname "$(find . -name "talent_graph.py" -type f 2>/dev/null | head -1)")
echo "Script directory: $SKILL_DIR"
Step 3: Verify
cd "$SKILL_DIR" && python3 talent_graph.py search "transformer" --limit 3
You should see a list of scholars.
GitHub Token (optional, strongly recommended): Enables automatic GitHub engineering data in profiles and raises API limit from 60/hour to 5000/hour.
export GH_API_TOKEN="your_personal_access_token"
Required inputs
- Search: English research topic, institution, paper area, or scholar name.
- Profile: scholar name; GitHub or X/Twitter handle only when user provides it or it is clearly public and verified.
- Institution analysis: English institution name and optional limit.
- Latest papers: keyword query and optional recency/limit.
When not to use
- Do not use for recruitment pipeline scoring or hiring decisions; use
ai-talent-radar for recruitment-oriented workflows.
- Do not use for precise Chinese-name disambiguation unless the user provides stable identifiers such as ORCID, homepage, affiliation, or paper title.
- Do not infer private identity, employment intent, compensation, or contact details from public academic data.
Scenario Mapping
| User says | Agent executes |
|---|
| Search scholars in reinforcement learning | cd $SKILL_DIR && python3 talent_graph.py search "reinforcement learning" |
| Find LLM researchers with high h-index | cd $SKILL_DIR && python3 talent_graph.py search "large language model" --hmin 20 |
| Find Python-based Agent researchers | cd $SKILL_DIR && python3 talent_graph.py search "AI agent" --lang python |
| Profile @karpathy | cd $SKILL_DIR && python3 talent_graph.py profile "Andrej Karpathy" --github karpathy |
| Analyze DeepMind | cd $SKILL_DIR && python3 talent_graph.py institution "DeepMind" |
| Latest RLHF papers | cd $SKILL_DIR && python3 talent_graph.py latest "RLHF reinforcement learning human feedback" |
| Output JSON for further processing | Add --json to any command |
| Check scholar X/Twitter influence (followers/Bio) | Use any X/Twitter scraping tool to fetch profile for the scholar's handle |
| Check scholar's latest tweets | Use any X/Twitter scraping tool to fetch timeline for the scholar's handle |
Note: Data sources are primarily English academic databases. Search domestic institutions using English names (e.g., ByteDance / Tsinghua University).
Execution Guide
Search Scholars
cd $SKILL_DIR
python3 talent_graph.py search "multimodal large language model" --limit 10 --hmin 15 --lang python
--hmin: Minimum h-index threshold
--lang: Programming language filter (requires GH_API_TOKEN for GitHub data)
--limit: Max results (default 10, max 20)
Generate Scholar Profile
python3 talent_graph.py profile "Yann LeCun" --github ylecun
--github: Optional, provide GitHub username for engineering data
- Without it, the system auto-guesses lowercase/no-space name variants (marked "unverified")
Supplement X/Twitter Data (optional, no login needed)
After generating a scholar profile, if you know the scholar's X/Twitter handle, supplement community influence data:
⚠️ If direct access to x.com is blocked, set HTTP_PROXY environment variable. Unknown handles can often be found on scholar homepages or Google Scholar profiles.
Query Institution
python3 talent_graph.py institution "OpenAI" --limit 10
Track Latest Papers (arXiv)
python3 talent_graph.py latest "AI agent autonomous" --limit 5
Returns recent arXiv preprints with abstracts and PDF links. arXiv enforces 3-second rate limiting; expect 5-10 second wait.
Error Handling
| Error | Resolution |
|---|
| ModuleNotFoundError: requests | Run pip install requests |
| ImportError module not found | Confirm you cd $SKILL_DIR before running; don't call from other directories |
| OpenAlex returns empty results | Hint: No matching scholars found; try English keywords |
| GitHub API 403 / rate limit | Hint: GitHub requests exhausted; configure GH_API_TOKEN; academic data only shown |
| ORCID timeout | Graceful degradation: returns OpenAlex data only, notes "ORCID data unavailable" |
| All data sources fail | Clearly state: data sources temporarily unavailable, please retry later |
Data Sources
| Platform | Data Dimensions | Free Quota | Notes |
|---|
| OpenAlex | Scholar/Paper/Citation/Institution | Unlimited (reasonable rate) | Primary source |
| Semantic Scholar | h-index/Citations | 100 req/5min | Cross-validation supplement |
| ORCID | Scholar profile/Employment history | Unlimited | Identity supplement |
| arXiv | Preprint papers | Unlimited (3s interval) | latest command only |
| GitHub | Projects/Tech stack | 60/hour (no token) | Stable with token configured |
Compliance
- Only uses official free public APIs, respects all platform rate limits
- No scraping, no anti-bot circumvention
- Data is for academic research and talent evaluation reference only
Gotchas
⚠️ OpenAlex access may occasionally time out from certain regions → timeout=15s set, auto-retry once on failure, graceful degradation on second failure
⚠️ Semantic Scholar API rate limit 100 req/5min, batch queries trigger 429 → 3s sleep between requests; 60s backoff on 429
⚠️ GitHub API without Token is limited to 60 req/hour, batch querying multiple scholars exhausts quota → strongly recommend configuring GH_API_TOKEN; without Token, query at most 3 scholars
⚠️ ORCID search yields no results for Chinese names (database is primarily English) → hint users to use English full names (First Last) or ORCID ID directly
⚠️ arXiv enforces 3s request interval, consecutive calls trigger 429 → latest command caps at 5 papers per batch with 3s sleep between requests
⚠️ If cd $SKILL_DIR fails (path not found), running python3 talent_graph.py directly causes ImportError → must confirm path via Quick Start Step 2 first
Hard Stop
If the same tool call fails more than 3 times, stop immediately. List all failed approaches and reasons, mark "Manual intervention needed", and wait for human confirmation.
Changelog
7.0.0
- Added X/Twitter scholar influence dimension: supplement followers, Bio, tweet count after profile generation
- Added X/Twitter trigger words in description
- Tags: added
twitter, x
- Scenario mapping: added X query scenarios
6.0.0
- Synced H1 title version with frontmatter
5.0.0
- Single-line description (YAML compatible), added "not applicable" scenarios
- Added
tags
- Synced scripts/ to latest versions
- API endpoints, quotas, compliance integrated into SKILL.md
4.0.0
- Integrated OpenAlex, arXiv, Semantic Scholar, ORCID four data sources
- Differentiated academic research orientation from recruitment orientation (ai-talent-radar)
- Added institution query and latest paper tracking
- Added optional GitHub Token configuration
1.0.0–3.0.0
- Initial: OpenAlex + arXiv dual data sources
- Progressive expansion with Semantic Scholar, ORCID supplements