60 skills I've written for Claude — the ones I actually use every day on a live HBO show, in ComfyUI, and building my own tools. A skill is a Markdown file with YAML frontmatter that Claude auto-invokes when a request matches its trigger description. They live in ~/.claude/skills/ and load per-conversation.
20 skills covering shot tracking, vendor management, EDL pipelines, CO3 coordination, and weekly prep on DTF St. Louis.
8 skills for spinning up a new VFX project — guided wizard plus subsystem connectors for email, ShotGrid, Slack, and calendars.
Legacy dashboard kept warm for archival lookups on Righteous Gemstones S2.
8 skills that compose into end-to-end creative pipelines — brief, build, run, iterate.
Turn scripts into visuals before a camera rolls.
Two skills driving p5 Studio, the live visuals engine I built for performance use.
12 skills covering how I actually ship code with Claude — brainstorm, plan, test, debug, review, ship.
Skills around StockBot — my options-screening dashboard and its sell-decision logic.
One thin wrapper that gets used a lot.
Housekeeping and meta-skills — health checks, skill authoring, and inbox triage.
Three skills I actually leaned on this year.
I use this on DTF St. Louis every Monday. I say /vfx-episode-summary 102 and it pulls the last week of VFX emails from Gmail, scans the EDL database for shot status across all 136 shots in the episode, checks what CO3 received vs. what we delivered, flags any shots that regressed or are missing expected versions, and writes a one-page report for the supervisor call. Before this skill, that took three hours with Red Pen Tools, the EDL watcher, and a lot of tab-switching. Now it takes ninety seconds. The same skill is what made the VFX Dashboard's "War Room" view possible.
I describe what I want — "photo-real still of Cyd the cat at 3/4 profile, lit like a Nat Geo cover, Flux with a Cyd LoRA" — and /comfy-build pulls from my engine registry (Flux, Wan, Hunyuan3D, Kling, Minimax, DALL-E, Image-1, SAM2), picks the right base engine, wires the node graph including LoRA loaders and sampler settings, and writes a workflow JSON I can load straight into ComfyUI. The skill knows my custom nodes, my LoRAs, my typical resolution presets, and the tricks I use (two-pass pose control for characters, IPAdapter for style transfer). It means I stop thinking in nodes and start thinking in shots.
Garage Force needed before/after transformation videos for 115 real garage-floor jobs, with distinctive transition effects that didn't look like stock. I used /comfy-hunt to find candidate effect workflows (magic snaps, particle dissolves, power washes, aurora waves, fairy dust, freeze shatters), /comfy-build to adapt each into a SAM2-segmented + Kling-i2v pipeline that isolates just the floor, and /comfy-run to queue dozens of iterations across properties. The same 3-act structure (declutter → transform → beauty) got generated 100+ times. The skills composed cleanly — that's the whole point of the architecture.
Available for editing, supervising, and tool-building consulting.
Contact