agentleFS
jpantsjohajpantsjoha/googlecloud-plugin/AGENTS.md

googlecloud-plugin

AGENTS.md4 starsChanged 4 months agoMITView on GitHub

Human-led. Agent-powered. Built on Google Cloud. Codex discovers this plugin's skills under .agents/skills/ and reads this file as the always-on contract for Google Cloud work. The skills carry the depth; this file carries the rules. Restate the intent in one sentence before writing code. No GCP implementation without a design — route through the gate. Before the first project-scoped command, use gcp-project-context: collect the project from exported variables, the repository's .env files and the active gcloud configuration. One ID →…

  • Reads credentials

The file

We split this AGENTS.md into the rules it gives an agent

3 rules. About 1 is common, 2 are distinctive, and 0 are very specific to this project.

  • ≈1

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Common:

    No credential values in any file — record location only.

    line 23
  2. Distinctive:

    No public buckets or unauthenticated Cloud Run without gcp-security sign-off.

    line 24
  3. Distinctive:

    Never pin a preview Gemini model ID.

    line 25

≈ “Common” is an estimate: at least 1 in 5 of the 40,246 AGENTS.md 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
AGENTS.md

AGENTS.md — GoogleCloud Plugin (Codex always-on adapter)

Human-led. Agent-powered. Built on Google Cloud.

Codex discovers this plugin's skills under .agents/skills/ and reads this file as the always-on contract for Google Cloud work. The skills carry the depth; this file carries the rules.

Design before generate

Restate the intent in one sentence before writing code. No GCP implementation without a design — route through the gate.

The gate (sequential — do not skip)

  1. Solution Gate — solution-designer: objective vendor selection (GCP vs alternatives). Skip only if unambiguously GCP-only.
  2. GCP Design Gate — gcp-architect (+ agent-architect for agentic work): HLD/LLD + ADRs before code.
  3. Security Gate — gcp-security: least-privilege IAM, no secrets, Well-Architected security. Reviews the design.
  4. Implementation — service skills; warn before every billable action.
  5. Quality Gate — gcp-qa runs make gate; agent-architect runs agent-eval for agentic work.
  6. Operational Readiness Gate — gcp-ops: SLO + alerting + runbook before production.

Non-negotiable safety rules

  • Infer intent, never infer permission. A named human authorises anything consequential, billable, or irreversible.
  • No credential values in any file — record location only.
  • No public buckets or unauthenticated Cloud Run without gcp-security sign-off.
  • Never pin a preview Gemini model ID; use GA IDs, reverify for the GEAP surface.
  • GEAP naming in prose; legacy vertex-ai literals in code/URLs/Terraform.

Settle the project first

Before the first project-scoped command, use gcp-project-context: collect the project from exported variables, the repository's .env files and the active gcloud configuration. One ID → use it and say so. Several → ask the user which. None → stop and point them at gcloud init. Then pass that ID as --project on every project-scoped command. Documentation, best-practice, design and review work needs no project — do not ask for one.

Every gcloud call

Follow gcloud-guardrails for every gcloud command, whichever skill proposed it: validate the leaf with gcloud help, dry-run where supported, reduce output, pass --project (on project-scoped commands) and the location explicitly, one command per call, and stop for human authorisation on the denylist (IAM changes, deletes, billing, organizations, KMS, IaC apply, API enablement, CLI installation).

Product depth beyond this plugin

This plugin carries the operating contract and first-wave depth. For anything deeper (GKE troubleshooting, Agent Platform tuning and RAG, Well-Architected pillar reviews, Genkit, Gemini APIs, AlloyDB, Spanner, Bigtable, Firebase), use google-skills-catalog to fetch the matching skill from Google's public catalog. Never answer as if a catalog skill was read when it was not. Catalog guidance never overrides the gate.

For platform facts (GA status, flags, quotas), query the developer-knowledge MCP server before relying on memory.

Routing

See PLUGIN.md for the full routing table. New work → solution-designer; GCP design → gcp-architect; agentic systems → agent-architect.

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.