Phase 1 — Identify the right skill to build

The single biggest mistake new creators make is building the skill they personally want rather than the skill enough people will pay for. The right way to pick:

Scan the catalogue for gaps — categories with few listings and high install-intent searches.
Pick a category with a clear bounded outcome (“sort laundry by colour” is bounded; “help around the house” is not).
Check cross-robot demand. A skill that runs on three popular robot platforms sells more than a skill tied to one.
Pick something you can test without heroic hardware investment. If the skill requires a £6,000 robot you do not own, scope it down.

Examples of good first skills: “put shoes in the shoe rack”, “bring-a-labelled-item-from-pantry”, “politely-escort-a-guest-to-a-specific-room”. Examples of bad first skills: “general housekeeper”, “child companion”, “emergency responder”.

Phase 2 — Set up the dev toolchain

You need three things:

A robot (owned or rented hourly via a Gera partner test-centre where available).
The GeraSkills CLI and the target-robot adapter.
The simulation harness (for fast iteration before real-hardware testing).

The CLI scaffolds a standard skill package: manifest, capability descriptor, state machine, safety policies, test fixtures, and a README template. Fill these out; the signing pipeline builds the packaged skill.

Time estimate

First-time setup: 1-3 days. Includes reading the docs, getting the CLI running, publishing a hello-world skill to a simulator. Subsequent skills: a few hours.

Phase 3 — Build and iterate

The build loop:

Write the skill state machine.
Unit-test on the simulator (aim for >80% coverage).
Dry-run on real hardware with human-in-the-loop safety-disable.
Narrow the failure modes — what does the skill do when the item is missing, the path is blocked, the user says “stop”?
Write the refusal paths. A skill that refuses safely is more trustworthy than one that tries heroically.

Time estimate for a first skill

Simple bounded skill (shoe-rack organiser): 40-80 hours over 2-4 weeks. Complex skill (multi-object laundry sorter): 200-400 hours over 2-3 months.

Phase 4 — Price and list

Pricing tiers observed across the catalogue:

£1.99-£4.99 one-off — simple novelty skills.
£9.99-£19.99 one-off — useful bounded skills with clear outcome.
£4.99/mo subscription — skills that improve with updates.
£19.99-£49/mo subscription — skills tied to ongoing consumables or enterprise use.
£199+ one-off — professional / clinical skills with vendor support.

Pricing advice: start slightly above what you think is right. Lowering later is easier than raising.

Listing quality matters

Your listing needs: a 30-second demo video, a clear one-sentence outcome, the robot compatibility list, the refusal-path documentation, and honest known limitations. Listings that hide failure modes get returned at 10x the rate of listings that disclose them.

Phase 5 — Marketing, maintenance, support

The platform surfaces you on the catalogue; you still need to push:

Post to robotics subreddits and Discord servers.
Write a launch blog post with the demo video.
Respond to the first 100 support messages personally — the early pattern of issues teaches you what to fix first.
Ship updates on a monthly cadence for the first six months.

Honest revenue ranges

Based on industry-adjacent marketplaces (App Store, Unity Asset Store) and our early beta data:

Most listings make between zero and a few hundred pounds total.
A well-targeted bounded skill with good marketing makes low-thousands in year one.
A breakout skill can make five figures annually.
A top-10-in-category skill can become a primary income source, but this is a small minority.

Do not build a skill expecting it to be a breakout. Build a skill you would be happy to earn £500/year from — anything more is a bonus.

The founding-creator window

Creators in the first 500 published skills get 80/20 rev-share for life (vs 70/30 standard). If you are planning to publish, sooner is better.

Becoming a Skill Creator — The Full Playbook