agentleFS
dvmkitdvmkit/dvmctl/skills/build-dvm/SKILL.md

build-dvm

Skill0 starsChanged 32 days agoApache-2.0View on GitHub

Build and operate a Digital Vending Machine with dvmkit. Use when asked to turn an idea into a paid service, implement a capability, test payments or Lightning-funded credit, deploy or update a DVM, inspect earnings or health, recover a failed operation, or retire a service.

The file

We split this skill into the rules it gives an agent

7 rules. About 0 are common, 5 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈5

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Use dvmctl to manage the service and @dvmkit/sdk to implement its capabilities.

    line 10
  2. Distinctive:

    Before a remote change, confirm the service, organization, and environment from tool output.

    line 35
  3. Distinctive:

    Keep private keys, recovery words, payment credentials, and secret values out of chat, source, and logs.

    line 39
  4. Distinctive:

    Keep the service's price separate from the authorized test-spend cap.

    line 40
  5. Distinctive:

    When an outcome is uncertain, inspect the existing operation before retrying.

    line 41
  6. Distinctive:

    Check user authorization before actions that spend funds, change payment destinations, transfer control, or destroy data.

    line 43
  7. Very specific:

    Read structured output and follow its recovery guidance.

    line 49

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Build and operate a DVM

Use dvmctl to manage the service and @dvmkit/sdk to implement its capabilities. Use the caller CLI, dvm, when testing the service as a customer. Keep its test wallet separate from the builder's keys and normal caller state.

Choose the procedure

Identify the user's task and the service's current state. Read the matching file before acting. Load another only when the task crosses into its scope; an existing service does not need onboarding again.

TaskRead
Turn an idea into a working local capabilitynew-builder
Validate behavior or prove a paid calltesting, then only the chosen payment recipe
Build a job that prompts, charges mid-job, runs long, or has several capabilitieswork-shaped jobs
Store the DVM's own data in Postgresbuilder tables
Prepare or perform a deploymentdeploying
Inspect health or earnings, collect payouts, refund a completed job, update, or transfer a serviceoperating
Diagnose a failure or reconcile an uncertain payment or deployrecovery
Retain job content for support or a disputecontent holds
Export minimized financial evidence before database deletionfinancial export
Stop offering a service and close its outstanding obligationsretirement

Establish the context

Reuse the brief, decisions, and authorization already provided. Ask only for missing choices that affect the result, cost, access, or recoverability. An instruction to test locally does not authorize a production deployment or spending production funds.

Inspect the installed CLI version and the project's SDK dependency before selecting examples. Use the exact release and matching skill specified by the builder's setup. Do not upgrade an existing project merely to match an example. If versions disagree, establish the intended target before changing dependencies.

Before a remote change, confirm the service, organization, and environment from tool output. Names alone can identify the wrong service when an account belongs to several organizations.

Rules that apply to every procedure

  • Keep private keys, recovery words, payment credentials, and secret values out of chat, source, and logs. Use the documented secret-input mechanism. Never infer that a human has backed up recovery material.
  • Keep the service's price separate from the authorized test-spend cap. Verify the network and payment destination before moving value. A development server can still accept real funds.
  • When an outcome is uncertain, inspect the existing operation before retrying. A lost response does not prove that a payment or deployment failed. Continue through recovery when the tool cannot establish the result.
  • Preserve recovery records until the relevant obligation is resolved. Unknown refund state is not a zero balance, and deleting a service does not settle its debts. Use retirement before destructive cleanup.
  • Check user authorization before actions that spend funds, change payment destinations, transfer control, or destroy data. Existing authorization counts; do not ask for it again. Ask when the actual scope or consequence exceeds it.

The dvmkit team welcomes feedback at any time through dvmctl feedback send, especially about what was confusing, slow, or broken.

Use the result

Read structured output and follow its recovery guidance. Relay display and hint when they answer the user's question. Check each command's contract: logs and exports need not be JSON.

Close a task with the observed result and the evidence relevant to it: a local artifact, a verified paid job, a live deployment, or a confirmed payout. When a local result closes a paid-service brief without deployment authorization, follow the deployment question in new-builder. Distinguish completed work from pending or unknown state. Never report a payout from a revenue total alone.

For exact syntax and output, read the dvmctl reference. For handler interfaces, read the SDK reference. Use the docs index to retrieve other topics selectively.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.