agentsclimarketplace
Sign in

Ai billing

Skill commet-labs/skills/skills/ai-billing

Implement and diagnose billing for AI usage with Commet, including the installed @commet/ai-sdk integration or direct usage reporting from Node, Python, Go, Java, and PHP. Use the installed SDK documentation for middleware options, token fields, pricing, errors, and balance behavior instead of copied formulas or API inventories.From its SKILL.md

Install
npx -y skills add commet-labs/skills --skill ai-billing

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 3 stars3 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

3.0 KB, as published. Nobody here has run it

AI billing

Treat AI billing as a measured billing workflow. Exact middleware options, token fields, pricing inputs, errors, and SDK calls come from the installed documentation and generated types.

Resolve installed context

  1. Detect the installed Commet SDK, AI provider or gateway, model wrapper, and the code path that receives authoritative usage totals.
  2. Locate the SDK's installed docs/manifest.json, read its entrypoint, and follow the indexed AI token billing, usage, balance, and error documentation.
  3. For @commet/ai-sdk, also read its installed README and peer dependency versions.
  4. In Node projects, run commet doctor --output agent and resolve failed package, documentation, compatibility, API-version, or project-context checks.

Do not use token field names, cost formulas, model identifiers, middleware options, or error codes from this skill. They change with the installed contract and provider integration.

Implement

  1. Confirm which customer and billable feature own the usage.
  2. Confirm the product's configured consumption model and price source before writing tracking code. Do not silently convert between metered, credits, or balance behavior.
  3. Record usage only from the provider's final authoritative usage result. Preserve streaming behavior and the original model response.
  4. Use a caller-owned logical event identity and the installed contract's idempotency mechanism so retries cannot double charge.
  5. Define the expected application behavior when tracking fails. Do not silently claim billing succeeded when reporting failed.
  6. Keep pricing and margin policy in Commet configuration when the installed product documentation assigns ownership there; do not duplicate it in application code.

Guard effects

Code edits and local provider mocks are local. Creating or changing plans, features, prices, model mappings, balances, subscriptions, or usage events is a remote billing effect.

Before a remote effect, name the exact organization and sandbox or live mode. Use a passing PROJECT_CONTEXT_VALID doctor check in Node projects; otherwise require explicit project or user context. Verify billing changes in sandbox unless the user explicitly authorizes the named live effect.

Never print API keys, provider keys, prompts, completions, or full customer payloads while diagnosing token billing.

Verify

Run the application's formatter, compiler or typecheck, and tests. Exercise successful usage reporting, retry with the same logical event identity, the chosen tracking-failure behavior, and the configured exhaustion behavior against sandbox when remote verification is authorized.

Keep looking

Skills are one crate of 627,801. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.