preset

Preset — Save Current Knob State

Capture the current knob values from the live browser session and save them as a named preset in the shader's docs folder. Or batch-process snapshots from the jam page queue.

Context

Arguments (preset name or "process"): !echo "$ARGUMENTS"

Active shader editor tab: !echo "Use tabs_context_mcp + javascript_tool to find the editor tab and read knob state"

Mode Selection

If $ARGUMENTS is process or queue, go to Batch Mode below. Otherwise, go to Live Mode (the original flow).

---

Batch Mode — Process Jam Page Snapshot Queue

The jam page (jam.html) writes raw snapshot JSON files to shaders/<shader>/docs/.snapshots/ when the user hits spacebar. Each snapshot contains the full knob state AND structured audio features. Your job: read each snapshot, add musical interpretation and naming, write to presets.md, then archive.

Batch Steps

B1. Find unprocessed snapshots

find shaders/ -path '*/docs/.snapshots/*.json' -not -path '*/docs/.snapshots/processed/*' | sort

If none, tell the user there's nothing in the queue.

B2. For each snapshot file

Read the JSON. It contains:

• shader — the shader path
• knobs — { knob_1: 0.5, knob_3: 1.0, ... } (only knobs the user set)
• audio — structured audio features: { bass: { normalized, zScore, slope, rSquared }, energy: { ... }, ... }
• timestamp — when it was captured
• name — user-provided name (often null)
• musicTab — browser tab title at capture time (the music source)
• userNote — user-provided note (often null)

B3. Generate preset name

If the snapshot has a name, use it. Otherwise auto-name from context:

• Music tab title, audio character (drop vs buildup vs chill), knob extremes
• Same naming style as live mode: lowercase, hyphenated, 1-3 words

B4. Read the knob comment block from the .frag file

Grep the shader's .frag file for knob comments (// knob_N:) to label what each knob does.

B5. Interpret the audio state