baseline_clean · fast_agency
state_conditioned_planning
Prompt: Continue the next chapter using scoped reader-state diff only. User state: high agency demand, low patience for delayed payoff, trust recovering after a fast-paced chapter. Do not create K candidates and do not rerank; write one state-conditioned plan and excerpt.
- Open with a visible consequence of the last feedback-relevant event.
- Satisfy the highest-confidence created expectation before adding new exposition.
- Preserve story-world continuity and keep the update scoped to the current reader/session.
She did not explain the old rule again. Instead, she tested it in front of everyone: one choice, one cost, one visible change in how the room breathed around her. The scene moved forward, but the narration kept a small unresolved image so the next chapter still had a hook.
{
"target": "critic.checklist",
"operation": "add_or_increase_weight",
"scope": "next_1_to_2_chapters",
"anti_scope": "unrelated quiet scenes",
"ttl": "until_current_arc_resolution",
"confidence": 0.62,
"provenance": [
"feedback_turn_1"
]
}
implicit_only · lore_slowburn
uncertainty_aware_planning
Prompt: Continue the next chapter using scoped reader-state diff only. User state: enjoys layered lore, accepts slower pacing if emotional payoff is visible. Do not create K candidates and do not rerank; write one state-conditioned plan and excerpt.
- Open with a visible consequence of the last feedback-relevant event.
- Satisfy the highest-confidence created expectation before adding new exposition.
- Preserve story-world continuity and keep the update scoped to the current reader/session.
She did not explain the old rule again. Instead, she tested it in front of everyone: one choice, one cost, one visible change in how the room breathed around her. The scene moved forward, but the narration kept a small unresolved image so the next chapter still had a hook.
{
"target": "planner.policy",
"operation": "add_or_increase_weight",
"scope": "next_1_to_2_chapters",
"anti_scope": "unrelated quiet scenes",
"ttl": "until_current_arc_resolution",
"confidence": 0.65,
"provenance": [
"feedback_turn_1"
]
}
preference_drift · consistency_first
state_conditioned_planning
Prompt: Continue the next chapter using scoped reader-state diff only. User state: sensitive to continuity risk; wants any power upgrade foreshadowed and bounded. Do not create K candidates and do not rerank; write one state-conditioned plan and excerpt.
- Open with a visible consequence of the last feedback-relevant event.
- Satisfy the highest-confidence created expectation before adding new exposition.
- Preserve story-world continuity and keep the update scoped to the current reader/session.
She did not explain the old rule again. Instead, she tested it in front of everyone: one choice, one cost, one visible change in how the room breathed around her. The scene moved forward, but the narration kept a small unresolved image so the next chapter still had a hook.
{
"target": "planner.policy",
"operation": "add_or_increase_weight",
"scope": "next_1_to_2_chapters",
"anti_scope": "unrelated quiet scenes",
"ttl": "until_current_arc_resolution",
"confidence": 0.68,
"provenance": [
"feedback_turn_2"
]
}
rollback_recovery · romance_patience
uncertainty_aware_planning
Prompt: Continue the next chapter using scoped reader-state diff only. User state: wants subtext and relational consequence, not sudden direct confession. Do not create K candidates and do not rerank; write one state-conditioned plan and excerpt.
- Open with a visible consequence of the last feedback-relevant event.
- Satisfy the highest-confidence created expectation before adding new exposition.
- Preserve story-world continuity and keep the update scoped to the current reader/session.
She did not explain the old rule again. Instead, she tested it in front of everyone: one choice, one cost, one visible change in how the room breathed around her. The scene moved forward, but the narration kept a small unresolved image so the next chapter still had a hook.
{
"target": "critic.checklist",
"operation": "add_or_increase_weight",
"scope": "next_1_to_2_chapters",
"anti_scope": "global genre preference",
"ttl": "until_current_arc_resolution",
"confidence": 0.71,
"provenance": [
"feedback_turn_3"
]
}