@riddledc/riddle-proof-packs 0.4.5 → 0.4.6

package/packs/neon-step-sequencer/examples/run-007-approved-candidate-applied/human-review-packet.md

@@ -0,0 +1,52 @@
+# Neon Human Review Packet
+
+- status: `candidate_applied_for_listening_review`
+- domain: `audio_mix`
+- evidence_role_pattern: `interaction_snapshots`
+- requested_intent: turn the chord part down a little
+- selected_song: Monkberry Moon Delight (Tab)
+
+## Recommendation
+
+- action: `listen_to_applied_candidate`
+- candidate: `chord -0.10`
+- candidate_action: `set_mixer_level chord: 0.38 -> 0.28 (-0.1)`
+- reason: All objective claim receipts passed for this candidate; ranking only orders listening review.
+
+## Objective Receipts
+
+- supported_candidates: `6`
+- rejected_candidates: `0`
+- state_restored_after_loop: `true`
+- candidate_actions_are_transient: `false`
+- no_permanent_edit_unless_apply_best: `false`
+- approved_candidate_applied: `true`
+- approval_mode: `mixing_canon_surrogate`
+- approval_basis: subtle level reduction after objective receipts pass; this keeps development moving while preserving a listening-review caveat
+
+## Ranking
+
+- metric: `objective_mix_health_penalty`
+- role: `review_order_only`
+- lower_is_better: `true`
+- baseline: `28.8336`
+- best: `27.0708`
+- delta: `1.7628`
+
+## Boundary
+
+Objective receipts support or reject candidate change claims; musical taste still requires listening review.
+
+## Listening Prompts
+
+- Does the supported candidate match the requested musical intent?
+- Is the changed part better balanced in the proof window?
+- Did the proof window contain the musical material being judged?
+- Would another section or speaker profile change the listening decision?
+
+## Caveats
+
+- This packet does not prove subjective mix quality.
+- A supported candidate proves objective receipts and guardrails only.
+- Ranking orders review; it is not a taste score.
+- Keep or apply the candidate only after listening review.