Skip to main content

skill-beziehungen

The AI assistant Klacksy learns from everyday usage which of its capabilities (skills) belong together — and next time selects the right skills more precisely as a result.

How it works​

In the background, Klacksy observes which skills are frequently used together or in sequence within a session and derives relations from that. The learning process runs automatically every 6 hours. The Skill Relations page shows all detected relations; administrators decide whether they are accepted or discarded.

Relation types

  • Sequential: Skill A is regularly used before skill B.
  • Jointly required: Both skills frequently occur in the same session, with no fixed order.

Source

  • Learned: Computed from real usage data — highly reliable.
  • Derived: Logically inferred from the structure of the skills, not yet confirmed by usage data.

Status

  • Candidate: Newly detected, awaiting your decision.
  • Active: Accepted — actively influences Klacksy.
  • Retired: Discarded — will be ignored.

The confidence (0–100%) shows how sure Klacksy is about a relation. It rises with every session that confirms the pattern (confirmations) and falls with sessions that speak against it (contradictions).

Good to know​

  • Accept a relation when it makes sense on the merits — the two skills genuinely belong together and the pattern matches how you want to work.
  • Discard relations that arose by chance or don't belong together in your context (e.g. from test sessions).
  • Rule of thumb: only learned relations with high confidence are typically worth accepting. Derived candidates with little support are better discarded until more usage data is available.

Try it yourself: Klacks Playground — login admin@test.com / P@ssw0rt1, data resets daily.