generated-content-quality
Installation
SKILL.md
Generated Content Quality
Purpose: Stop a content corpus from collapsing into three shapes wearing forty costumes.
Sources: The saturation-ledger method here is generalised from a working
implementation in the Royal Inbox project (docs/CONTENT_NOVELTY_LEDGER.md),
which measures structural saturation across a 47-event corpus and sets explicit
quotas for the next pack. The underlying convergence problem is documented in
the LLM-diversity literature (mode collapse under RLHF; Kirk et al., "Understanding
the Effects of RLHF on LLM Generalisation and Diversity," 2024). No published
game-industry standard exists for this — the method below is a practice, not a
citation, and its thresholds are posture B under the Numbers Policy.