github-trending

name

github-trending

description

Generate a daily GitHub trending project-intelligence report: fetch trending repos, study README/API evidence, answer each repo's purpose/content/value in Chinese, and publish a lens-first HTML digest to html-serve. Use when the user asks for GitHub trending, hot repos, 今日 GitHub 热门, trending 项目报告, or a daily open-source project digest. Do NOT use for deep research on a single repo (use deep-research) or general AI news briefs (use daily-ai-brief).

GitHub Trending

Pipeline: Fetch → Study → Analyze → Render → Publish

Output Contract

This skill exists to answer one daily question: 今天的热门项目到底在解决什么问题、做了什么、创造了什么价值？

For every repo, produce a Chinese three-lens analysis:

Lens	Required answer	Evidence to use
项目目的	It solves what problem, for whom, in what context.	README intro, tagline, topics, homepage, repo description.
项目内容	It actually provides what product, library, workflow, model, app, or infrastructure.	README features, install/usage, directory/docs when cloned.
项目价值	It creates what practical value: speed, cost, quality, access, automation, developer leverage, or ecosystem signal.	Use cases, adoption signals, stars-today, comparisons, integration surface.

Do not treat stars as value by themselves. Stars are attention; value must be stated as a user-facing or ecosystem-facing gain.

Workflow

Stage 1: Fetch

omp github-trending fetch --since daily --out /tmp/trending.json

Options: --since daily|weekly|monthly, --lang <language>, --readme-chars N. Output: JSON array, one object per repo with name, desc, lang, stars_today, stars, forks, topics, license, created_at, readme. Repos that failed enrichment carry an error field instead.

Done when: JSON file exists with 10+ repos, or the hard gate explains why it cannot.

Stage 2: Study

Read the JSON before writing. Use readme, full_desc, topics, homepage, language, age, and stars-today to infer the three lenses.

Only clone when the three lenses are still unclear and the repo ranks in the top 4:

gh repo clone <owner>/<repo> /tmp/<repo> -- --depth 1

Inspect README, docs, examples, package metadata, and top-level structure. Never clone more than 3 repos.

Done when: every repo has evidence for purpose/content/value, or an explicit uncertainty note.

Stage 3: Analyze

For each repo, write these fields in Chinese:

Field	Length	Rule
purpose	1 sentence	Name the pain point or job-to-be-done.
content	1 sentence	Say what the repo contains or enables, not only what category it belongs to.
value	1 sentence	State the practical gain or strategic signal.
trend_reason	0-1 sentence	Explain why it may be trending today when evidence supports it.
summary	1 sentence	A compact editorial judgment tying the three lenses together.

Then write 今日看点: one paragraph that names the day's dominant problem cluster, the strongest value signal, and the top mover.

Done when: every repo answers all three lenses; uncertain claims are marked as uncertain instead of invented.

Stage 4: Render

Choose one template from assets/ based on the report's reading goal. If the user names a template, use it. Otherwise choose freely.

Template	Use when	Required fragments
report-template.html	The report should read like an editorial digest with a strong lead story.	{{FOCUS_REPOS}}, {{REST_REPOS}}
signal-dashboard-template.html	The report should emphasize trend clusters, signal strength, and fast scanning.	{{TREND_MAP}}, {{REPO_SIGNAL_CARDS}}

Sort repos by stars-today descending. Keep the same analysis obligation regardless of template:

Rank	Required emphasis
1	purpose/content/value + editorial judgment.
2-4	compact three-lens analysis.
5+	scannable three-lens rows or cards.

Replace all {{MARKER}} placeholders. Generate escaped HTML fragments using the structure documented in the chosen template comments. HTML-escape repo descriptions, README-derived text, topics, and analysis.

Write the rendered HTML to a temporary local file. Filename convention: <YYYY-MM-DD>.html (weekly/monthly runs: <YYYY-MM-DD>-<since>.html).

Also write the analyzed data to a sibling <same-name>.json so later agents can query the report structurally instead of parsing HTML:

{
  "date": "<YYYY-MM-DD>",
  "since": "daily",
  "highlight": "<今日看点 paragraph>",
  "repos": [
    {
      "name": "owner/repo", "url": "https://github.com/owner/repo",
      "lang": "...", "stars_today": 0, "stars": 0, "topics": ["..."],
      "desc": "...", "purpose": "...", "content": "...", "value": "...",
      "trend_reason": "...", "summary": "..."
    }
  ]
}

Keep repos in the same stars-today order as the HTML.

Done when: local HTML and JSON files exist, no {{ markers remain in the HTML, and every repo in the JSON carries purpose/content/value.

Stage 5: Publish

Publish through html-serve CLI:

omp html-serve publish <tmp.html> --to github-trending/<filename>.html --source github-trending --tag github --tag trending
omp html-serve publish <tmp.json> --to github-trending/<filename>.json --source github-trending --tag github --tag trending --tag data --title "GitHub Trending <YYYY-MM-DD> data"

Tell the user both returned URLs of the HTML: localhost_url and tailscale_url.

Done when: both omp html-serve publish calls succeed and the HTML URLs are given to user.

Hard Gate

Condition	Action
fetch returns no repos	Abort; tell user to check network or report page-structure breakage.
gh not authenticated	Continue; the script falls back to anonymous API with 60 req/h. Suggest gh auth login for higher limits.
More than half of repos carry error	Warn user about API rate limit; suggest gh auth login or retry later.
A repo's purpose/content/value cannot be inferred	Mark the specific lens as 证据不足; do not hallucinate. Clone only if it is top 4 and clone budget remains.
omp html-serve publish reports missing HTML_SERVE_DATA_DIR	Ask user to configure docker/html-serve/.env or export it; do not write to a hardcoded path.
html-serve container not running	Tell user: omp container up html-serve.
Template file missing	Abort; tell user to reinstall the skill.

Command Hierarchy

omp
└── github-trending
    └── fetch --since daily|weekly|monthly --lang <language> --readme-chars N --out <path>