CRUX-Compress: Skill for Cursor

CRUX Skill: Memory Meditation — Research Protocol

Implements the full Research-mode meditation contract: Phases A–G recursive depth-first exploration, depth-0 manager steps 1–13 (including step 4b 4-mode additional_focus_areas[] reconciliation, step 8 K10c in-pass reflection, and step 8b respawn-payload prep), facet registry lock, citations index, peer review file spec, and comprehensiveness honouring at leaf depth.

When to Use

Load this skill when:

• The crux-cursor-meditation-guide agent is spawned in Research mode (no --quick flag, meditateMode: "research")
• You are executing the depth-0 manager workflow (steps 1–13)
• You are executing as a child Research-mode agent at depth 1, 2, or 3 (Phases A–G)
• You are executing as a peer-review agent (peerReviewForBranch set)

Never load this skill in Quick mode, Ensemble Aggregation mode, or Adversarial Review function — those are owned by meditation-quick, meditation-ensemble, and meditation-review respectively.

Prerequisites

1. Verify flags.enableMemories is "true" (abort if not)
2. Confirm comprehensiveness: payload is present in the spawn prompt — abort with: "comprehensiveness: payload required; missing from spawn prompt — caller misconfigured" if missing
3. Confirm theming: payload is present — abort with a clear error pointing at Theme Preflight if missing
4. Confirm modelStrategy: payload is present — abort with: "modelStrategy: payload required; missing from spawn prompt — caller misconfigured" if missing

Input Parameters (depth-0 manager invocation)

The depth-0 manager receives these parameters from the calling agent:

meditateMode: "research"
workingDir: "<absolute path to meditations/{yyyymmdd}-{topic-slug}/>"
theming: { ... }                  # Theme Preflight payload — REQUIRED
comprehensiveness: { ... }        # Richness payload — REQUIRED; abort if missing
modelStrategy: { ... }            # Model strategy payload — REQUIRED; abort if missing
confirmDeepFacets: "none" | "depth_2_only" | "all_levels"
maxDepth: 1 | 2 | 3               # default 3
ensembleModel: "<model-slug>"     # derived from modelStrategy at the depth-0 manager; carrier for per-spawn model: selection
preConfirmedFacets: [ ... ]       # only when in ensemble_max — skip step 4 derivation