Fact preferences and LLM-based tagging

This OpenClaw skill fetches well-sourced facts from Wikipedia's Did You Know? section — one at a time, no repeats, no API key required. Install from ClawHub or see the README for manual setup.

What's new

Preferences

• Tell your agent what you enjoy ("more history, less science") and facts are scored and ranked accordingly.
• Two dimensions: domain (20 tags covering topic areas like history, music, science, animals) and tone (9 tags like quirky, dark, inspiring, surprising).
• Facts are tagged automatically when the cache refreshes, so scoring stays current.

Automatic scoring signals

• Newer facts get a freshness boost.
• Consecutive same-domain facts are penalised to keep variety.
• Shorter, punchier hooks are slightly preferred.
• Facts with multiple source links get a small bonus.

Scheduled delivery

• Ask the agent to deliver facts on a schedule ("send me a fact every morning at 8am").
• Works independently of preferences — with or without them.

Prefs CLI

• prefs.py provides direct management of preferences: init, list, get, set.

Under the hood

• Module restructure: dyk.py split into helpers.py (utilities), serve_hook.py (refresh and serving logic), and a thin backwards-compatible shim.
• Cache expiry switched from collection-count limit to age-based (8-day rolling window).
• Score breakdown dicts (candidate_score, served_score) persisted to cache for LLM explainability.
• Comprehensive test coverage across all modules (test_helpers, test_serve_hook, test_dyk, test_fetch_hooks, test_write_tags, test_prefs).

Bug fixes and hardening

• Adds cooldown throttle to prevent hammering the Wikipedia API on repeated invocations
• Fixes atomic cache writes via temp-file-and-rename to prevent corruption
• Fixes URL encoding storage and deduplication so percent-encoded and plain URLs are treated as the same article
• Fixes deduplication history so trimming old collections cannot re-expose previously seen facts
• Fixes HTML entity decoding,   handling, and bold-italic markup stripping in fact text
• Fixes graceful handling of unreadable or malformed cache files
• Adds test suite

Initial release to ClawHub

• Initial release of the "did-you-know" skill, available on ClawHub https://clawhub.ai/jonathandeamer/did-you-know
• Fetches English Wikipedia’s "Did you know?" facts.
• No API key required.
• Facts are cached locally and served one at a time.
• Includes clear responses for no new facts and error cases.
• Does not modify Wikipedia content.