- name
- github-rising-stars
- description
- Discover fast-growing GitHub repositories before they appear on the official Trending page. Use when users ask to find emerging/rising GitHub projects, spot early-stage repos gaining momentum, identify hidden gems not yet on Trending, or discover the hottest new repos in the last 1-7 days. Triggers on phrases like "find rising GitHub repos", "GitHub trending before trending", "fast growing repos", "GitHub黑马项目", "增速快的GitHub项目", "还没上trending的项目", "发现早期项目".
GitHub Rising Stars
Discovers fast-growing GitHub repositories that haven't made the official Trending page yet, by querying the GitHub Search API across multiple dimensions and ranking by stars per day.
Quick Start
Run the bundled script directly:
python3 scripts/find_rising.pyDefault: repos created in the last 3 days, min 20 stars, top 15 results.
Script Options
python3 scripts/find_rising.py [options]
--days N Look-back window in days (default: 3, range: 1-7)
--min-stars N Minimum star count threshold (default: 20)
--top N Number of results to return (default: 15)
--lang LANG Filter by programming language (e.g. Python, TypeScript, Go)
--json Output machine-readable JSONExamples
# Find Python repos rising fast in the last 5 days
python3 scripts/find_rising.py --days 5 --lang Python
# Find top 20 repos with at least 50 stars
python3 scripts/find_rising.py --min-stars 50 --top 20
# JSON output for further processing
python3 scripts/find_rising.py --jsonHow It Works
- Multi-query search — runs 3 GitHub API queries with overlapping time/star ranges to maximize coverage
- Deduplication — merges results by repo ID
- Spam filtering — removes known low-quality signals (crypto bots, game hacks, follower farms, etc.)
- Rate calculation — computes
stars / age_in_daysfor each repo - Ranked output — sorts by rate descending, shows top N
Interpreting Results
- Rate > 200/day: Viral — likely hits Trending within 24h
- Rate 50–200/day: Fast-rising — worth watching
- Rate < 50/day: Early signal — may accelerate
Repos with no description or no language tag may be spam or private mirrors — apply extra scrutiny.
Agent Workflow
When a user asks to find rising repos:
- Run
python3 scripts/find_rising.pywith appropriate flags - Present results in a clean table, grouped by theme/category if patterns emerge
- Highlight repos worth deeper investigation (high rate + meaningful description + known author)
- Optionally open top repos in browser for quick review
Limitations
- GitHub Search API is unauthenticated by default — rate limit: 10 requests/min. Add a
Authorization: token <PAT>header in the script for higher limits. - Results reflect repos *created* recently; established repos that suddenly spike are not captured (use
pushedfilter variant for that). - Stars can be artificially inflated; cross-reference with Hacker News / Reddit for validation.