Skip to content

Using it from an AI agent

SKILL.md in the repository root is an agent skill (the same shape as Archify's). Install it with npx skills add akashtalole/archify-gcp, then ask your agent, for example:

Use archify-gcp to diagram a multi-zone web app on GKE with Cloud SQL, estimate the monthly cost, and review it against the Well-Architected pillars.

What the agent should do

  1. Choose the diagram type with archify-gcp guide "<scenario>".
  2. Find icons with archify-gcp icons search <term>; never invent ids.
  3. Write the spec following the Google Cloud authoring rules (structure first, two zones or regional tiers for HA, full product name once).
  4. Add usage to nodes if the user gave volumes; otherwise leave defaults and say the estimate is indicative.
  5. Run finalize --json and repair until it passes (about four rounds at most), then report what remains.
  6. Look at the PNG. The receipt says visualReview: "not-performed".
  7. Report the absolute paths of the outputs, node and edge counts, warnings, the monthly cost with its confidence and key assumptions, the findings by risk, how many recommendations were actually evidenced, and the sentence "cost is an indicative list-price estimate and the review is advisory — inferred from the drawing, not from deployed configuration".

Rules the skill enforces on the agent

Cost (rules adapted from the AWS billing-and-cost-management skill)

  • Check today's date before quoting prices; the price book records its retrieval date.
  • Never do cost arithmetic by reasoning. Run archify-gcp cost and quote its numbers.
  • Public pay-as-you-go rates only. Never invent a price; unpriced components stay not-itemized or not-estimated.
  • Point to the Google Cloud Pricing Calculator for a quote.

Well-Architected (structure adapted from the AWS well-architected-review skill)

  • The corpus is the frozen Google Cloud Well-Architected Framework snapshot; use canonical recommendation ids only (REL-3.1, SEC-1.4…).
  • Five statuses; say Cannot Determine rather than guess.
  • Do not manufacture Critical findings; acknowledge strengths.
  • The report is confidential: do not post it in broadly visible channels without approval.

Machine-readable output

Pass --json whenever the agent parses a result. schema <type> returns the JSON Schema for the spec, and validate --json returns { ok, errors, warnings }.

archify-gcp validate spec.json --json
archify-gcp finalize spec.json --json
archify-gcp cost spec.json --json
archify-gcp wa review spec.json --json