Pick and generate scene images for a session — spawns the scene-image-generator sub-agent which reads the session HTML, recommends cinematic moments, generates landscape PNGs, and wires them into the HTML. DM approves each image via SendMessage before it is inserted.
Parse all source files in campaigns/[name]/sources/ and shard them into the campaign directory structure — encounters, characters, locations, and state. Run once before the first session.
Verify and display the full encounter block for a named encounter, then prepare for combat. Always cross-checks stat blocks against the SRD before displaying.
Distribute loot from the last encounter — prompts who takes what, handles currency splitting, tracks magic item attunement, and updates all character files.
Format all scene narration, NPC dialogue, and DM guidance consistently — read-aloud text the DM speaks at the table, separated from DM-only context Claude provides as insight. Apply automatically during any D&D session narration, NPC interaction, location description, or scene setup.
Recovery skill — rebuilds the session conversation log from the raw JSONL transcript when the normal incremental flow failed or context was lost. Use only when the conversation log is missing or incomplete.
Walk through a short or long rest for one or more characters — hit dice rolls, HP recovery, feature restoration, spell slot recovery. Updates all character files.
First-time setup and add-campaign — creates shared website infrastructure on first run, then walks through setting up a campaign folder. Deploys CLAUDE.md, .gitignore, and JS/CSS from templates. Safe to run multiple times.
Update framework files (CLAUDE.md, .gitignore, JS, CSS) from the .claude templates after pulling fixes from the repo. Never touches campaign content, config, or the root website/index.html.
★not rated 3 4d agoA44 tokens
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: