Project Radar
A daily trending & community discovery radar, packaged as a Claude Agent Skill — it pulls the full GitHub Trending list and scans what's discussed across Hacker News, Reddit, Product Hunt, Google, YouTube, X and the wider web, scores each candidate 0–100 on real-world usefulness, renders a single self-contained Turkish HTML report in a fixed signature design, and keeps a persistent watchlist with per-project detail reports.
🇹🇷 Türkçe için README.tr.md
Quick Start
git clone https://github.com/sametbrr/project-radar.git
cp -r project-radar ~/.claude/skills/project-radar
pip install jinja2 markupsafe
Start a new session, then trigger it with project radar çalıştır / run the radar (or invoke /project-radar).
Features
- Discovery-only — never clones, installs, builds, or runs candidate repos; it reads sources, evaluates, and reports.
- Usefulness scoring — scores each candidate 0–100 on real-world usefulness; AI/LLM tooling is a prominent interest but not the only filter.
- Full GitHub Trending — a deterministic scraper (
fetch_github_trending.py) pulls the entire daily Trending list across all languages, not a fixed subset; weekly/monthly on demand.
- Multi-source scan — Hacker News, Reddit, Product Hunt, Google, YouTube, X and the wider web; general tech + product agenda, not only AI/LLM.
- Persistent watchlist — add any GitHub repo; its README + metadata are read into a medium-depth detail report (Markdown + signature HTML) and listed back with file references.
- Signature HTML report — one self-contained Turkish HTML file in a fixed design (single Inter font, animations on); PDF optional.
- Agent-agnostic — a standard
SKILL.md folder that works in any agentskills.io-compatible tool.
Requirements
- Python 3 — for the report renderer and helper scripts.
jinja2 and markupsafe — required for HTML output.curl — used by the Trending scraper and watchlist fetch (falls back to Python urllib).gh CLI — optional; the watchlist uses it for authenticated GitHub API, falling back to unauthenticated curl.playwright + Chromium — optional, only for PDF output.- An agent that supports the
SKILL.md format (Claude Code, Hermes, Codex CLI, Cursor, Gemini CLI …).
Installation
The skill is a standard SKILL.md folder, so it works in any agent that supports the format.
# Personal (all projects) — Hermes runs on top of Claude Code, so this also
# makes the skill available to Hermes on the same machine.
git clone https://github.com/sametbrr/project-radar.git
cp -r project-radar ~/.claude/skills/project-radar
# project-scoped instead: cp -r project-radar .claude/skills/project-radar
# Runtime deps (HTML output needs only the first two):
pip install jinja2 markupsafe
# Optional PDF output:
pip install playwright && playwright install chromium
For Claude.ai / Cowork: Settings → Capabilities → Skills → upload the packaged project-radar.skill.
Start a new session and run /skills to confirm it loaded.
Usage
Render the bundled example to verify everything works, then use the helper scripts directly.
# Render the bundled example:
python3 scripts/build_report.py \
--data assets/report-data.example.json \
--out Project-Radar-$(date +%F).html
# Pull the full GitHub Trending list (deterministic, all languages):
python3 scripts/fetch_github_trending.py --out /tmp/github-trending-raw.json
# Watchlist: add a repo, render its detail report, list everything with refs:
python3 scripts/watchlist.py add owner/repo
python3 scripts/watchlist.py render owner__repo
python3 scripts/watchlist.py list
# Mobile-friendly PDF instead (A4 portrait, paginated):
python3 scripts/build_report.py --data <data.json> --out report.pdf
Configuration
PROJECT_RADAR_WATCHLIST_DIR — where the persistent watchlist lives (default: watchlist/ under the skill directory). The watchlist.py --dir flag overrides it.- The report design — colors, score thresholds, single Inter font — is fixed in
assets/report-template.html by design and is not meant to be edited.
How It Works
fetch_github_trending.py ─┐
HN · Reddit · PH · Google · YouTube · X · web ─┤→ agent scores & filters → report-data.json → build_report.py → HTML
