Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Search 2M+ education research records via the ERIC database API
.claude/skills/brycewang-stanford-eric-education-api/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 87% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 73% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 93% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 39% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 41% | 0% |
ERIC is the world's largest digital library of education research, sponsored by the U.S. Institute of Education Sciences (IES). It indexes 2M+ records including journal articles, reports, conference papers, and dissertations covering all aspects of education. The API provides free, unauthenticated access to metadata and links to full text where available.
https://api.ies.ed.gov/eric/bash# Basic keyword search curl "https://api.ies.ed.gov/eric/?search=online+learning&format=json&rows=20" # Search in specific fields curl "https://api.ies.ed.gov/eric/?search=title:\"blended learning\"&format=json" # Filter by publication date curl "https://api.ies.ed.gov/eric/?search=STEM+education&start=0&rows=25&\ publicationdatestart=2023-01-01&publicationdateend=2026-12-31&format=json" # Filter by publication type curl "https://api.ies.ed.gov/eric/?search=formative+assessment&\ publicationtype=Journal+Articles&format=json" # Filter by descriptor (ERIC thesaurus term) curl "https://api.ies.ed.gov/eric/?search=descriptor:\"Higher Education\"&format=json" # Peer-reviewed only curl "https://api.ies.ed.gov/eric/?search=metacognition&peerreviewed=true&format=json"
| Parameter | Description | Example | |-----------|-------------|---------| | search | Free-text or field search | search=adaptive+learning | | format | Response format | json or xml | | rows | Results per page (max 200) | rows=50 | | start | Pagination offset | start=50 | | publicationtype | Document type | Journal Articles, Reports, Dissertations/Theses | | publicationdatestart | From date | 2024-01-01 | | publicationdateend | To date | 2026-12-31 | | peerreviewed | Peer-reviewed filter | true or false | | descriptor | ERIC thesaurus term | descriptor:"Distance Education" | | educationlevel | Education level | Higher Education, Elementary Education | | subject | Subject area | subject:"Mathematics Education" |
| Field | Description | |-------|-------------| | title | Article title | | author | Author name | | descriptor | ERIC controlled vocabulary term | | source | Journal/source name | | abstract | Abstract text | | id | ERIC document ID (e.g., EJ1234567) |
| Type | Description | |------|-------------| | Journal Articles | Peer-reviewed journal articles | | Reports - Research | Research reports | | Reports - Descriptive | Descriptive reports | | Reports - Evaluative | Program evaluations | | Dissertations/Theses | Graduate research | | Speeches/Meeting Papers | Conference presentations | | Books | Books and book chapters |
json{ "response": { "numFound": 8450, "start": 0, "docs": [ { "id": "EJ1389012", "title": "Effects of AI Tutoring on Student Learning Outcomes", "author": ["Smith, John", "Chen, Wei"], "source": "Journal of Educational Technology", "publicationdateyear": 2024, "description": "This study examines the impact of AI-powered tutoring...", "descriptor": ["Artificial Intelligence", "Tutoring", "Academic Achievement"], "educationlevel": ["Higher Education"], "peerreviewed": "T", "url": "https://eric.ed.gov/?id=EJ1389012", "publicationtype": "Journal Articles", "issn": "1234-5678" } ] } }
pythonimport requests BASE_URL = "https://api.ies.ed.gov/eric/" def search_eric(query: str, rows: int = 25, peer_reviewed: bool = True, pub_type: str = None, from_year: int = None) -> list: """Search the ERIC education research database.""" params = { "search": query, "format": "json", "rows": rows, } if peer_reviewed: params["peerreviewed"] = "true" if pub_type: params["publicationtype"] = pub_type if from_year: params["publicationdatestart"] = f"{from_year}-01-01" resp = requests.get(BASE_URL, params=params) resp.raise_for_status() data = resp.json() results = [] for doc in data.get("response", {}).get("docs", []): results.append({ "id": doc.get("id"), "title": doc.get("title"), "authors": doc.get("author", []), "source": doc.get("source"), "year": doc.get("publicationdateyear"), "abstract": doc.get("description", "")[:300], "descriptors": doc.get("descriptor", []), "level": doc.get("educationlevel", []), "url": doc.get("url"), }) return results def search_by_descriptor(descriptor: str, rows: int = 50) -> list: """Search using ERIC thesaurus controlled vocabulary.""" return search_eric(f'descriptor:"{descriptor}"', rows=rows) # Example: find recent AI in education research papers = search_eric("artificial intelligence classroom", from_year=2023, rows=10) for p in papers: print(f"[{p['year']}] {p['title']}") print(f" Descriptors: {', '.join(p['descriptors'][:5])}") # Example: search by ERIC descriptor papers = search_by_descriptor("Gamification") for p in papers: print(f"{p['id']}: {p['title']} — {p['source']}")
ERIC uses a controlled vocabulary of 12,000+ descriptors for consistent indexing. Key descriptors include:
| Descriptor | Coverage | |------------|----------| | Distance Education | Online/remote learning | | Educational Technology | EdTech tools and methods | | Higher Education | University-level education | | STEM Education | Science, technology, engineering, math | | Teacher Education | Teacher training and development | | Assessment | Testing and evaluation | | Curriculum Development | Curriculum design | | Special Education | Inclusive education |
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-09 | pass→pass | 9,442 | 2,137 | -77% | 1 | 1 | 0% | 1,362 | 2,113 | +55% | 0 | 0 | — |
case-18 | pass→pass | 8,953 | 2,931 | -67% | 1 | 1 | 0% | 1,521 | 2,135 | +40% | 0 | 0 | — |
case-01 | fail→pass | 12,155 | 17,187 | +41% | 1 | 1 | 0% | 2,137 | 3,988 | +87% | 0 | 0 | — |
case-02 | fail→pass | 13,294 | 10,330 | -22% | 1 | 1 | 0% | 1,926 | 3,334 | +73% | 0 | 0 | — |
case-03 | fail→pass | 9,292 | 4,577 | -51% | 1 | 1 | 0% | 1,392 | 2,680 | +93% | 0 | 0 | — |
case-04 | pass→pass | 7,829 | 5,288 | -32% | 1 | 1 | 0% | 1,486 | 2,719 | +83% | 0 | 0 | — |
case-05 | fail→pass | 10,852 | 2,913 | -73% | 1 | 1 | 0% | 1,669 | 2,321 | +39% | 0 | 0 | — |
case-06 | fail→pass | 11,925 | 4,702 | -61% | 1 | 1 | 0% | 1,800 | 2,535 | +41% | 0 | 0 | — |
case-07 | pass→pass | 8,482 | 4,172 | -51% | 1 | 1 | 0% | 1,480 | 2,522 | +70% | 0 | 0 | — |
case-08 | pass→pass | 7,927 | 3,137 | -60% | 1 | 1 | 0% | 1,248 | 2,288 | +83% | 0 | 0 | — |
case-10 | pass→pass | 7,724 | 4,726 | -39% | 1 | 1 | 0% | 1,282 | 2,445 | +91% | 0 | 0 | — |
case-11 | pass→pass | 4,775 | 2,316 | -51% | 1 | 1 | 0% | 676 | 2,070 | +206% | 0 | 0 | — |
case-12 | fail→pass | 12,865 | 10,448 | -19% | 1 | 1 | 0% | 2,081 | 3,549 | +71% | 0 | 0 | — |
case-13 | pass→pass | 7,719 | 5,235 | -32% | 1 | 1 | 0% | 1,494 | 2,699 | +81% | 0 | 0 | — |
case-14 | fail→pass | 4,507 | 1,309 | -71% | 1 | 1 | 0% | 794 | 1,895 | +139% | 0 | 0 | — |
case-15 | pass→pass | 6,316 | 2,211 | -65% | 1 | 1 | 0% | 865 | 2,207 | +155% | 0 | 0 | — |
case-16 | pass→pass | 4,955 | 3,681 | -26% | 1 | 1 | 0% | 721 | 2,300 | +219% | 0 | 0 | — |
case-17 | pass→pass | 10,383 | 7,176 | -31% | 1 | 1 | 0% | 1,773 | 2,885 | +63% | 0 | 0 | — |
case-19 | pass→pass | 8,045 | 2,924 | -64% | 1 | 1 | 0% | 1,300 | 2,264 | +74% | 0 | 0 | — |
case-20 | pass→pass | 4,483 | 6,535 | +46% | 1 | 1 | 0% | 791 | 2,707 | +242% | 0 | 0 | — |
case-21 | pass→pass | 7,331 | 7,798 | +6% | 1 | 1 | 0% | 1,342 | 3,395 | +153% | 0 | 0 | — |
case-22 | pass→pass | 3,269 | 2,566 | -22% | 1 | 1 | 0% | 518 | 2,171 | +319% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +32 percentage points is the difference between those two pass rates over the 22 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.