Skip to content

frankzch/ai-news-skill

ai-news-skill

Verified

Fetch the latest AI/ML news, trending open-source projects, social media discussions, and KOL (Key Opinion Leader) perspectives from 370+ curated sources. Use when the user asks about recent AI news, AI trends, what's happening in machine learning, open-source AI projects, AI Twitter/Reddit discussions, or wants a daily AI briefing.

AI Intelligence Skill

When to use this skill

Use when the user wants to:

  • Get the latest AI/ML news and announcements
  • Browse trending open-source AI projects from GitHub
  • Read AI discussion highlights from Reddit, X (Twitter), HackerNews
  • See what AI KOLs (Key Opinion Leaders) are saying
  • Get a comprehensive daily AI briefing

Workflow

Default: Fetch latest AI briefing

Run the fetcher script from the skill directory:

uv run scripts/pulse_fetcher.py

If uv is not available, install dependencies manually and use:

pip install requests pyyaml
python scripts/pulse_fetcher.py

Filtered fetch

Map the user's intent to CLI flags:

User saysFlags
"Only Reddit discussions"--include-categories discussion --include-sources reddit
"Open source, no Github"--include-categories opensource --exclude-sources github
"50 items, past 48 hours"--limit 50 --hours 48
"Show full summaries with links"--show-long-summary true --show-link true
"AI news in Chinese"--include-categories news --language zh

Available options

Run uv run scripts/pulse_fetcher.py --help for the full list. Key flags:

  • --include-categories "cat1,cat2" — Only fetch these categories: news, opensource, discussion, kol
  • --exclude-categories "cat1,cat2" — Skip these categories
  • --include-sources "src1,src2" — Only from these sources
  • --exclude-sources "src1,src2" — Skip these sources
  • --hours INT — Time window in hours (default: 24)
  • --limit INT — Optional cap on the total number of items (default: no extra cap). Each category is additionally capped by tier (guest top 5 / free top 10 / member all).
  • --language en|zh — Output language (default: auto-detect from system locale)
  • --show-short-summary true|false — Show short summary (default: true)
  • --show-long-summary true|false — Show long summary (default: false)
  • --show-link true|false — Show article URLs (default: false)
  • --output-file FILE — Export full JSON to file (default: pulse_output.json). Pass empty string to disable.

Output format

Each item is printed as:

1. [NEWS] Title of the article (from TechCrunch)
   Short Summary: One-line summary text
----------------------------------------

When --output-file is used, the full JSON response is saved to disk. See references/API.md for the response schema.

Gotchas

  • HTTP 429 means the user hit an IP rate limit (anti-spam interval or daily cap). Tell them to wait a few minutes and try again.
  • Tier caps per category. The response includes tier and tier_limit (per-category cap, null = unlimited); each category returns at most:
    • guest (no / invalid api_key): the top 5 items
    • free (logged in, non-member): the top 10 items
    • member: all items If the user asks for more items than the tier allows, the server silently clamps. To raise the cap, tell the user to register at InBrief.info, open Settings → PulseAI Agent Skill, copy the API Key, and paste it into the api_key field of config.yaml (see assets/config.default.yaml).
  • Use --output-file when links are included — terminal line-wrapping can corrupt long URLs.
  • If config.yaml exists in the skill root, it overrides default settings. See assets/config.default.yaml for the template.
  • The --include-* flags take precedence over --exclude-* flags for the same dimension.
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 322kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 322kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 322kUpdated Claude CodeCodexVerifiedInstall