Before you start

You need: a robot you own or have test access to, a Stripe- supported country for payouts, basic comfort with the command line, and ideally some experience in the robotics stack (ROS, the robot-specific SDK, or similar). You do not need to be a PhD robotics researcher — the SDK abstracts a lot.

Step 1: create your creator account

Sign up at skills.gera.services/create. Complete identity verification (passport or national ID). Connect a Stripe Connect account — this is where payouts land. The founding-creator 80/20 rev-share is only available during the first 500 published skills, so there is a mild first-mover premium.

Step 2: install the SDK

Install the GeraSkills CLI and the target-robot adapter. A standard skill scaffold gives you:

manifest.toml — metadata, permissions, compatibility.
src/ — behaviour code (Python or Rust for v1).
models/ — any trained weights or calibration data.
tests/ — simulated-environment tests.
safety/ — declared-safety checks and fail-safes.

Step 3: write the skill

Keep the behaviour scoped. A skill that does one thing well sells better than one that does five things badly. Be explicit about the environment assumptions (lighting, clutter, specific objects) — buyers reward honest scoping and penalise overclaiming.

Step 4: test

The SDK includes a simulator that runs your skill under a range of conditions. Submissions fail review if they do not include pass results on the standard test harness. Beyond the simulator, run on physical hardware — simulation-to-real is the hard part.

Step 5: declare permissions and safety

List every actuator your skill moves, every sensor it reads, every environment region it operates in, and every network endpoint it calls. Be tight: a skill that claims permissions it doesn’t need gets a worse trust badge.

Declare fail-safes. "If the skill detects a human in the action zone within 30cm, halt and wait." "If the skill loses network, finish the current step and idle." The platform inspects these declarations and the safety team reviews them manually.

Step 6: submit for safety review

Submission triggers: automated static analysis, simulator regression, human safety reviewer inspection, and (for high-risk category skills) a physical test-bench validation. Review time is typically 3-10 business days. Reviewer feedback is specific; iterate.

Step 7: price and publish

Set your price. Free, one-off (common range £0.99-£49), or subscription. Pick an early access option if you want founding-buyer discount and an early-access badge. Write the listing: title, summary, what the skill does, what it doesn’t do, compatibility, screenshots or a demo video.

Step 8: payouts

Stripe Connect pays out on your schedule (daily / weekly / monthly, minimum £25 balance). The 70/30 (or 80/20 for founding creators) is computed per transaction; refunds clawback pro- rata. Tax handling is yours; we issue year-end reporting.

After publish

Support your users. Skills with responsive creators retain higher ratings. Update your skill as robot firmware evolves — outdated compatibility is a common reason for listings to decay.

What sells

Narrow, daily-use skills: morning coffee, making the bed, simple yard work, pet-feeding.
Accessibility skills: fetch items for users with limited mobility.
Small-business skills: shelf restocking, delivery drop-off, warehouse sorting.

Common rejection reasons

Overclaimed capability — the listing promises more than the skill delivers.
Missing fail-safes for skills that move near humans.
Model files not under creator’s licence to distribute.
Skill behaviour that bypasses declared permissions.

How to Publish Your First Robot Skill and Earn From Downloads